JP2023140853A - Computer program, information processing device and method - Google Patents

Abstract

To provide a computer program, an information processing device and a method which are used to estimate a health condition of an object user on the basis of a history of behaviors executed by the object user and have at least partially improved performance.SOLUTION: A method makes at least one processor function so as to output object sleeping state data including at least one of the object sleeping state data for identifying a sleeping state of an object user, which are object quality data for identifying sleeping quality of the object user, object rhythm data for identifying sleeping rhythm of the object user, object time data for identifying a sleeping time of the object user and object category data for identifying a sleeping category to which the object user belongs from an estimation model by acquiring object behavior data for identifying a history of behaviors executed by the object user and inputting the object behavior data to the estimation model generated by executing supervised learning.SELECTED DRAWING: Figure 3

Description

The technology disclosed in this application identifies the state of health of a user who is the target of estimation (target user) by inputting data that identifies the history of actions performed by the user into a learning model. The present invention relates to a computer program, an information processing device, and a method for outputting data from the learning model.

Japanese Patent No. 6916367 (Patent Document 1) discloses a method of constructing an estimation model for estimating the health condition of a target user based on data describing the purchase history of the target user. Note that this patent document is incorporated in its entirety into the present specification by reference.

Patent No. 6916367

BACKGROUND OF THE INVENTION In recent years, there has been a need to provide a method with improved performance as a method for estimating the health status of a target user based on the history of actions performed by the target user.

Therefore, the technology disclosed in the present application provides a computer program that is used to estimate the state of health of a target user based on the history of actions performed by the target user, and has at least partially improved performance. , provides an information processing device and method.

A computer program according to one aspect is "generated by being executed by at least one processor to obtain target behavior data that identifies a history of actions performed by a target user, and performing supervised learning." Target sleep state data that identifies the sleep state of the target user by inputting the target behavior data into an estimation model, which includes target property data that identifies the sleep quality of the target user, and sleep of the target user. Target sleep state data including at least one of target rhythm data that identifies the rhythm of the target user, target time data that identifies the sleep time of the target user, and target category data that identifies the sleep category to which the target user belongs. The at least one processor may be operated to output from the estimation model.

An information processing device according to one embodiment includes “at least one processor, the at least one processor acquires target behavior data that identifies a history of actions performed by a target user, and performs supervised learning. Target sleep state data that identifies the sleep state of the target user by inputting the target behavior data into an estimation model generated by the method, and target property data that identifies the sleep quality of the target user; At least one of target rhythm data that identifies the sleep rhythm of the target user, target time data that identifies the sleep time of the target user, and target category data that identifies the sleep category to which the target user belongs. The estimation model may be configured to output target sleep state data including target sleep state data.

According to one aspect, a method is a method performed by at least one processor that executes computer-readable instructions, the at least one processor executing computer-readable instructions to identifying the sleep state of the target user by inputting the target behavior data into an estimation model generated by performing supervised learning; The target sleep state data includes target property data that identifies the sleep quality of the target user, target rhythm data that identifies the sleep rhythm of the target user, target time data that identifies the sleep time of the target user, and , and target sleep state data including at least one of the following: , target category data identifying a sleep category to which the target user belongs, from the estimation model.

According to another aspect, a method is a method performed by at least one processor executing computer-readable instructions, the at least one processor, by executing the instructions, The data includes sample behavior data identifying the history of actions performed by one sample user corresponding to the group, and sample sleep state data identifying the sleep state of one sample user corresponding to the group. an acquisition step of acquiring a plurality of sets of training data, and inputting the plurality of sets of training data into a learning model to learn the target behavior data, which identifies the history of actions performed by the target user. Target sleep state data that is input to identify the sleep state of the target user, target property data that identifies the sleep quality of the target user, target rhythm data that identifies the sleep rhythm of the target user; An estimation model configured to generate target sleep state data including at least one of target time data that identifies the sleep time of the target user and target category data that identifies the sleep category to which the target user belongs. ,
and a generation step for generating.

A computer program according to another aspect is “generated by being executed by at least one processor to obtain target behavior data that identifies a history of actions performed by a target user, and performing supervised learning. Target menopausal state data that identifies the state of menopausal symptoms of the target user by inputting the target behavioral data into an estimation model, the target comprehensive state identifying the comprehensive state of menopausal symptoms of the target user. Target menopausal state data including at least one of data, target first state data that identifies the target user's first symptom state, and target second state data that identifies the target user's second symptom state. The at least one processor may be operated to output from the estimation model.

An information processing device according to another aspect includes “at least one processor, the at least one processor acquires target action data that identifies a history of actions performed by a target user, and executes supervised learning. By inputting the target behavior data into the estimation model generated by At least one of target comprehensive state data for identification, target first state data for identifying the state of the first symptom of the target user, and second target state data for identifying the state of the second symptom of the target user. The estimation model may be configured to output target menopausal state data including one from the estimation model.

According to another aspect, a method is a method performed by at least one processor that executes computer-readable instructions, the at least one processor executing computer-readable instructions that The state of menopausal symptoms of the target user is determined by inputting the target behavior data into an estimation model generated by performing supervised learning. The target menopausal state data to be identified includes target comprehensive state data that identifies the comprehensive state of menopausal symptoms of the target user, target first state data that identifies the state of the first symptom of the target user, and outputting target menopausal state data including at least one of target second state data identifying the state of the second symptom of the target user from the estimation model.

According to another aspect, a method is a method performed by at least one processor executing computer-readable instructions, the at least one processor, by executing the instructions, The data includes sample behavior data that identifies the history of actions performed by one sample user corresponding to the group, and sample menopausal status data that identifies the state of menopausal symptoms of one sample user corresponding to the group. and target behavior data that identifies a history of actions performed by the target user by inputting the plurality of sets of training data into a learning model and causing it to learn. is the target menopausal state data that identifies the state of the menopausal symptoms of the target user, the target comprehensive state data that identifies the comprehensive state of the menopausal symptoms of the target user, and the first symptom of the target user. configured to output target menopausal state data including at least one of target first state data that identifies the state of the target user, and target second state data that identifies the state of the second symptom of the target user. and a generation step of generating an estimated model.

FIG. 1 is a block diagram showing an example of the configuration of a communication system according to an embodiment. FIG. 2 is a block diagram showing an example of the hardware configuration of the server device 10 shown in FIG. 1. As shown in FIG. FIG. 3 is a flow diagram showing an example of operations performed by the communication system 1 shown in FIG. FIG. 4 is a diagram showing an example of the Athens Insomnia Scale. FIG. 5A is a diagram illustrating an example of the Pittsburgh Sleep Questionnaire. FIG. 5B is a diagram illustrating an example of the Pittsburgh Sleep Questionnaire. FIG. 5C is a diagram illustrating an example of the Pittsburgh Sleep Questionnaire. FIG. 6A is a diagram showing an example of a three-dimensional sleep scale. FIG. 6B is a diagram showing an example of a three-dimensional sleep scale. FIG. 7 is a diagram showing an example of an insomnia severity questionnaire. FIG. 8A is a diagram showing an example of the Munich chronotype questionnaire. FIG. 8B is a diagram showing an example of the Munich chronotype questionnaire. FIG. 9 is a diagram showing an example of sleep category data used in the communication system 1 shown in FIG. 1. FIG. 10 is a diagram showing another example of sleep category data used in the communication system 1 shown in FIG. 1. FIG. 11 is a diagram conceptually showing an example of a search table used in the communication system 1 shown in FIG. 1. FIG. 12 is a flow diagram showing another example of operations performed by the communication system 1 shown in FIG. 1. FIG. 13 is a diagram showing an example of the simplified menopause index (SMI). FIG. 14 is a diagram conceptually showing another example of the search table used in the communication system 1 shown in FIG. 1. FIG. 15 is a diagram schematically showing the flow of data for predicting a customer's health using the health prediction system according to the embodiment. FIG. 16 is a diagram showing a specific example of the functional configuration of the integrated health prediction program 100 according to the embodiment. FIG. 17 is a diagram showing a specific example of purchase data according to the embodiment. FIG. 18 is a diagram showing a specific example of purchase data according to the embodiment. FIG. 19 is a diagram showing a specific example of a table for realizing a proposal to a customer according to an output value of a prediction model. FIG. 20 is a diagram showing a specific example of a table for realizing a proposal to a customer according to an output value of a prediction model. FIG. 21 is a diagram showing the processing flow of the integrated health prediction program according to the embodiment. FIG. 22 is a diagram illustrating a specific example of the hardware configuration of an information processing device on which the integrated health prediction program according to the embodiment is executed. FIG. 23 is a diagram illustrating a specific example of a functional configuration of a learning system including a model learning program according to an embodiment. FIG. 24 is a diagram showing a specific example of product categories. FIG. 25 is a diagram showing a specific example of the relationship between menopausal symptoms and health issues found from items purchased by subjects in the H group regarding menopausal symptoms. FIG. 26 is a diagram showing a specific example of a product category for women with menopausal symptoms who belong to the H group. FIG. 27 is a diagram showing the processing flow of the model learning program according to the embodiment. FIG. 28 is a diagram illustrating a specific example of a hardware configuration of an information processing device on which a model learning program according to an embodiment is executed.

Hereinafter, various embodiments of the present invention will be described with reference to the accompanying drawings. Note that common components in the drawings are given the same reference numerals. Also, it should be noted that components depicted in one drawing may be omitted from another drawing for convenience of explanation. Furthermore, it should be noted that the attached drawings are not necessarily drawn to scale.

The various systems, methods, and apparatus described herein are not to be construed as limited in any way. In fact, this disclosure covers each of the various disclosed embodiments, combinations of these various embodiments with each other, and combinations of portions of these various embodiments with each other. All novel features and aspects are directed. The various systems, methods, and apparatus described herein are not limited to the particular aspects, particular features, or combinations of particular aspects and particular features described herein. The articles and methods described herein do not require that one or more particular advantages be present or that a problem be solved. Furthermore, various features or aspects, or portions of such features or aspects, of the various embodiments described herein may be used in combination with each other.

The operations of some of the various methods disclosed herein are described in a particular order for convenience; It should be understood to include rearranging the order of the above operations unless otherwise required by the text. For example, operations listed in order are sometimes reordered or performed simultaneously. Furthermore, for purposes of brevity, the accompanying drawings do not depict the various ways in which the various matter and methods described herein may be used in conjunction with other matter and methods.

Any theory of operation, scientific principle, or other theoretical description presented herein in connection with the disclosed apparatus or method is provided for the purpose of better understanding and to limit the technical scope. not intended. The apparatus and methods in the appended claims are not limited to apparatus and methods that operate in accordance with such theories of operation.

Any of the various methods disclosed herein may be implemented using a plurality of computer-executable instructions stored on one or more computer-readable media and further executed in a computer. can be done. The one or more media may be non-transitory computer readable storage media, such as at least one optical media disk, volatile memory components, or non-volatile memory components. obtain. Here, the plurality of volatile memory components include, for example, DRAM or SRAM. Further, the plurality of nonvolatile memory components include, for example, a hard drive and a solid state drive (SSD). Moreover, the computer includes any computer available on the market, including, for example, smart phones and other mobile devices that have hardware to perform calculations.

Any of a plurality of such computer-executable instructions for implementing the techniques disclosed herein may be generated and used during implementation of the various embodiments disclosed herein. It may be stored along with any data on one or more computer readable media (eg, non-transitory computer readable storage media). Such computer-executable instructions may be part of a separate software application, for example, or may be accessed or downloaded via a web browser or other software application (such as a remote computing application). It may be part of a software application that is Such software may be installed, for example, on a single local computer (e.g., as a process running on any suitable computer available on the market) or in a networked environment using one or more networked computers. (eg, the Internet, a wide area network, a local area network, a client-server network (such as a cloud computing network), or other such network).

For clarity, only certain selected aspects of the various software-based implementations are described. Other details well known in the art are omitted. For example, the techniques disclosed herein are not limited to any particular computer language or program. For example, the techniques disclosed herein may be implemented by software written in C, C++, Java, or any other suitable programming language. Similarly, the techniques disclosed herein are not limited to any particular computer or type of hardware. Specific details of suitable computers and hardware are well known and need not be described in detail herein.

Furthermore, various such software-based embodiments (e.g., including computer-executable instructions for causing a computer to perform any of the various methods disclosed herein) Any of the above may be uploaded, downloaded, or accessed remotely by any suitable communication means. Such suitable communication means include, for example, the Internet, the World Wide Web, intranets, software applications, cables (including fiber optic cables), magnetic communications, electromagnetic communications (including RF communications, microwave communications, infrared communications), Including electronic communications or other such means of communication.

I. Chapter 1
1. Overview In the technology disclosed in the present application, to put it simply, at least one information processing device acquires target behavior data that identifies the history of actions performed by a user who is the target of estimation (target user). By inputting this target behavior data into the estimation model, data for identifying the health state of the target user (health state data) can be outputted from the estimation model.

Furthermore, the at least one information processing device may determine at least one target product and/or at least one target service corresponding to the health status data of the target user from among the multiple products and/or multiple services. can. Furthermore, at least one information processing device may output information identifying at least one target service and/or at least one target service determined in this way.

2. Configuration of Communication System Such a technique can be realized using a communication system as illustrated in FIG. FIG. 1 is a block diagram illustrating an example of the configuration of a communication system according to an embodiment.

As shown in FIG. 1, a communication system 1 according to an embodiment includes, for example, at least one server device (information processing device) 10 connectable to a communication network 2, and at least one terminal connectable to the communication network 2. device (information processing device) 20. In FIG. 1, two server devices 10A and 10B are illustrated as at least one server device 10, but any number of server devices 10 may be used. Similarly, although two terminal devices 20A and 20B are illustrated as at least one terminal device 20 in FIG. 1, any number of server devices 10 may be used.

Each server device 10 is connectable to at least one other server device 10 and/or at least one terminal device 20 via the communication network 2. Further, each terminal device 20 is connectable to at least one other terminal device 20 and/or at least one server device 10 via the communication network 2.

Each server device 10 may be any information processing device. Such information processing devices can include, for example, without limitation, personal computers, workstations, supercomputers, mainframes, and the like. Each terminal device 20 may be any information processing device. Such information processing devices can include, for example, without limitation, mobile phones, smartphones, tablets, personal computers, workstations, personal digital assistants, and the like.

In one example, at least one server device 10 provides a service for estimating the health status of any user ("estimation service") and/or at least one target product corresponding to the health status of any user. It may be an information processing device managed by an operating company that operates a proposed service (“proposed service”). Note that hereinafter, for convenience, the term "estimation service, etc." may be used to refer to the estimation service and/or proposal service.

In one example, at least one server device 10 and/or at least one terminal device 20 may be an information processing device managed by a corporate user and/or individual user who receives estimation services and the like from the operating company.

The operation of acquiring the target behavior data described above may be executed by at least one server device 10 and/or at least one terminal device 20.

The estimation model described above may be generated by at least one server device 10 and/or at least one terminal device 20. The estimated model generated in this way may be held by at least one server device 10 and/or at least one terminal device 20. The estimated model held in this way can be used by at least one server device 10 and/or at least one terminal device 20.

The operation of outputting the health state data described above from the estimation model may be performed by at least one server device 10 and/or at least one terminal device 20.

The operation of determining at least one target product and/or at least one target service described above may be performed by at least one server device 10 and/or at least one terminal device 20. The above-described operation of outputting at least one target service and/or data identifying at least one target service may be performed by at least one server device 10 and/or at least one terminal device 20.

Note that the communication network 2 includes a mobile phone network, a wireless network, a fixed telephone network, the Internet, an intranet, a local area network (LAN), a wide area network (WAN), and/or an Ethernet (registered trademark) network. including without limitation. The wireless network may be limited to, for example, RF connectivity via Bluetooth, WiFi (such as IEEE 802.11a/b/n), WiMax, cellular, satellite, laser, or infrared. can be included without exception.

3. Hardware Configuration of Each Information Processing Device Next, an example of the hardware configuration of each information processing device (server device 10, terminal device 20) will be described.

(1) Hardware configuration of the server device 10 FIG. 2 is a block diagram showing an example of the hardware configuration of the server device 10 shown in FIG. (described in connection with the terminal device 20).

As shown in FIG. 2, the server device 10 includes a central processing unit 11, a main storage device 12, an input/output interface device 13, an input device 14, an auxiliary storage device 15, and an output device 16. I can do it. These devices are connected to each other by a data bus and/or a control bus.

The central processing unit 11 is referred to as a "CPU" and can perform calculations on instructions and data stored in the main storage device 12 and store the results of the calculations in the main storage device 12. Furthermore, the central processing unit 11 can control the input device 14, the auxiliary storage device 15, the output device 16, etc. via the input/output interface device 13. The server device 10 may include one or more such central processing units 11.

The main storage device 12 is called a "memory" and stores instructions and data received from the input device 14, the auxiliary storage device 15, the communication network 2, etc. (terminal device 20, etc.) via the input/output interface device 13, and The calculation results of the central processing unit 11 can be stored. The main storage device 12 includes volatile memory (e.g., registers, cache, random access memory (RAM)), non-volatile memory (e.g., read-only memory (ROM), EEPROM, flash memory), and storage (e.g., hard disk drive). computer readable media such as, but not limited to, hard drives (HDDs), solid state drives (SSDs), magnetic tape, optical media). As will be readily understood, the term "computer-readable recording medium" refers to a transmission medium such as a modulated or transitory data signal, and not a transmission medium such as a modulated data signal or a transitory signal, but rather a data storage medium such as a memory and a storage medium. A medium can be included.

The auxiliary storage device 15 is a storage device that has a larger capacity than the main storage device 12. The auxiliary storage device 15 stores instructions and data (computer programs) constituting a specific application, and is controlled by the central processing unit 11 to input and output these instructions and data (computer programs). It can be transmitted to the main storage device 12 via the interface device 13. The auxiliary storage device 15 can include, but is not limited to, a magnetic disk device and/or an optical disk device.

The specific applications referred to here include an operating system, various software applications (including web browsers), dedicated applications executed to provide estimation services, etc., and dedicated applications executed to receive estimation services, etc. This may include, but is not limited to, applications such as:

The input device 14 is a device that takes in data from the outside, and can include, but is not limited to, a keyboard, a touch panel, a button, a mouse, and/or a sensor (microphone, camera), and the like.

The output device 16 can include, but is not limited to, a display device, a touch panel, a printer device, and the like.

In such a hardware configuration, the central processing unit 11 sequentially loads instructions and data (computer programs) constituting the specific application stored in the auxiliary storage device 15 into the main storage device 12, and It is possible to operate on the commands and data that have been created. Thereby, the central processing unit 11 controls the output device 16 via the input/output interface device 13, or controls other devices (for example, another server device 10) via the input/output interface device 13 and the communication network 2. and/or the terminal device 20, etc.).

In this way, the server device 10 performs operations related to providing estimation services and/or receiving estimation services, etc. (FIGS. 3 and 3) by executing the above-mentioned installed specific application. 12, etc., as described below). In addition to or in place of this, the server device 10 executes a browser and accesses other server devices 10 to provide and/or receive estimation services, etc. Related operations (such as various operations described below with reference to FIGS. 3, 12, etc.) can also be performed.

Note that the server device 10 can also include one or more microprocessors and/or graphics processing units (GPUs) instead of or together with the central processing unit 11.

(2) Hardware configuration of terminal device 20 As shown in parentheses in FIG. 2, the terminal device 20 can have substantially the same hardware configuration as the server device 10. As shown in FIG. 2, the terminal device 20 includes a central processing unit 21, a main storage device 22, an input/output interface device 23, an input device 24, an auxiliary storage device 25, and an output device 26. I can do it. These devices are connected to each other by a data bus and/or a control bus. Each of these devices is as described in relation to the server device 10.

The specific applications stored in the auxiliary storage device 25 and executed by the central processing unit 21 include an operation system, various software applications (including a web browser), dedicated applications executed to receive provision of estimation services, etc. These may include, but are not limited to, the following. Also, the specific application may include a dedicated application that is executed to provide an estimation service or the like.

In such a hardware configuration, the central processing unit 21 sequentially loads instructions and data (computer programs) constituting the above-mentioned specific application stored in the auxiliary storage device 25 into the main storage device 22, and It is possible to operate on the commands and data that have been created. Thereby, the central processing unit 21 controls the output device 26 via the input/output interface device 23, or controls other devices (for example, the server device 10, etc.) via the input/output interface device 23 and the communication network 2. / or other terminal devices 20, etc.) various information (data) can be sent and received.

In this way, the terminal device 20 executes the installed specific application to receive the estimation service, etc. and/or perform operations related to providing the estimation service, etc. (FIGS. 3 and 3). 12, etc., as described below). In addition to or in place of this, the server device 10 executes a browser and accesses other server devices 10 to provide and/or receive estimation services, etc. Related operations (such as various operations described below with reference to FIGS. 3, 12, etc.) can also be performed.

Note that the terminal device 20 can also include one or more microprocessors and/or graphics processing units (GPUs) instead of or together with the central processing unit 21.

4. Operations performed by communication system 2 to provide sleep state data regarding sleep state Next, regarding specific examples of operations performed by communication system 2 described above to provide sleep state data regarding sleep state. , will be further explained with reference to FIG. FIG. 3 is a flow diagram showing an example of operations performed by the communication system 1 shown in FIG.

(1) Step 1000
First, in step (hereinafter referred to as "ST") 1000, an information processing device, for example, a server device 10A managed by a company that operates an estimation service, etc. Teacher data can be acquired and stored.

Each set of teacher data that constitutes a plurality of sets of teacher data includes data that identifies the history of actions performed by one sample user corresponding to this set ("sample action data"), and data that identifies the history of actions performed by one sample user corresponding to this set. Data identifying the sleep state of one sample user ("sample sleep state data"). For example, the first set of teacher data includes sample behavior data that identifies the history of actions performed by one sample user (Mr. A) corresponding to the first set, and a sample that identifies the state of sleep of Mr. A. The second set of teacher data includes sample behavior data that identifies the history of actions performed by one sample user (Mr. B) corresponding to the second set; sample sleep state data identifying the sleep state of the person.

(1A) Sample Behavior Data Each sample behavior data may be data that identifies at least one (one or more) behaviors out of a plurality of predetermined behaviors. Here, each of the plurality of predetermined actions may be an action that can at least partially have a positive or negative impact on the state of sleep of the person who performed the action. Such a plurality of predetermined actions can include, for example, without being limited to, those exemplified below.
(Example of actions that can at least partially positively influence the state of human sleep)
・Performed aerobic exercise for at least 1 hour at least 2 days a week.
-Woke up at a certain time every day.
・Limited drinks containing caffeine to no more than 2 drinks per day.
・Working hours per day were 8 hours or less.
Purchased 10 or more bottles of coffee (this behavior can be identified, for example, by data including the product category to which the coffee belongs and the quantity purchased).
(Examples of actions that can at least partially negatively affect the state of human sleep)
・Performed aerobic exercise for 1 hour or more no more than once a week.
-Woke up at a different time every day.
・Drinked three or more drinks containing caffeine in a day.
・Working hours per day were 13 hours or more.
Purchased 5 or more bottles of milk (this behavior can be identified, for example, by data including the product category to which the milk belongs and the quantity purchased).

Each sample behavior data may be data that identifies at least one behavior among the plurality of predetermined behaviors illustrated in this way.

In addition, in one example, the products purchased by the user and/or the services used by the user may tend to have a positive or negative impact on the user's sleep state, so the plurality of predetermined actions may be , the product purchased by the sample user (or the category to which the product belongs), and the service used by the sample user (or the category to which the service belongs). This type of behavior can be generated using, for example, POS (Point of Sales) data held by retail stores (drug stores, supermarkets, convenience stores, etc.).

In one example, each of the plurality of predetermined behaviors is assigned unique identification data (such as letters, numbers, or a combination thereof), and each sample behavior data may include such identification data. can. For example, when using POS data, each of the plurality of sample behavior data includes JICFS classification (classification code), product category, purchase amount, purchase quantity, number of purchases, gender, age, etc., corresponding to the purchased product. , and/or date of purchase, etc., without limitation.

The server device 10A, which is an information processing device, for example, managed by a company that operates an estimation service, etc., can acquire such sample behavior data, for example, by at least one method among the methods exemplified below. can.
- The information processing device can access at least one other server device 10 (for example, another server device 10 installed in a drug store etc. that stores POS data, or this POS data) via the communication network 2. POS data is received from another server device 10).
- The information processing device sends, via the communication network 2, an email or a web page containing answers to a predetermined questionnaire (questionnaire regarding the behavior performed by the sample user) to at least one terminal device 20 and/or at least one Received from server device 10.
- The information processing device receives the contents of the responses via the communication network 2 from at least one server device 10 that has collected responses from a plurality of users to the above-mentioned predetermined questionnaire.
- The information processing device receives the contents of the responses to the above-mentioned predetermined questionnaire via a recording medium (USB memory, DVD-ROM, etc.) that has collected responses from a plurality of users.

(1B) Sample sleep state data The sample sleep state data included in each set of teacher data can include at least one of the data exemplified below, without being limited thereto.
・Data that identifies the sleeping characteristics of one sample user corresponding to the group (“sample characteristics data”)
・Data that identifies the sleep rhythm of one sample user corresponding to the group (“sample rhythm data”)
・Data that identifies the sleeping time of one sample user corresponding to the group (“sample time data”)
・Data that identifies the sleep category to which (the sleep of) one sample user corresponding to the group belongs ("sample sleep category data")

First, the sample property data may be data that identifies whether the sample user's sleep quality is good (eg, how good it is) or poor (eg, how bad is it). In one example, the sample property data may be data that identifies a score that overall evaluates the sleep quality of the sample user. In this case, the sample user's sleep quality may be, for example, data that identifies a score calculated by any of the methods exemplified below.

- A score calculated using the method described in the Athens Insomnia Scale, based on the sample user's answers to the questions listed in the Athens Insomnia Scale (AIS) (see Figure 4), or a score calculated using an arbitrary method. Corrected score obtained by performing calculations (addition, subtraction, multiplication, division, etc.) -Pittsburgh sleep questionnaire based on the sample user's answers to the questions listed in the Pittsburgh Sleep Questionnaire (PSQI) (see Figures 5A to 5C) A score calculated using the method described in the form, or a modified score obtained by calculating this score using any method (addition, subtraction, multiplication, division, etc.) - Three-dimensional sleep scale (3DSS) (Figures 6A and 6B A score calculated using the method described in the 3-dimensional sleep scale based on the sample user's answers to the questions described in (Reference), or calculating this score using any method (addition, subtraction, multiplication, division, etc.) - Based on the sample user's answers to the questions listed in the Insomnia Severity Questionnaire (ISI) (see Figure 7) The calculated score, or a modified score obtained by calculating this score using any method (addition, subtraction, multiplication, division, etc.) - Any other sample user's answers to the questions listed on any questionnaire or scale. a score calculated using the method described in the questionnaire or scale, or a modified score obtained by calculating this score using any method.

Second, the sample rhythm data may be data that identifies whether the quality of the sample user's rhythm is good (for example, how good it is) or bad (for example, how bad it is). In one example, the sample rhythm data may be data that identifies a score that evaluates the sleep rhythm of the sample user.

In this case, the sample user's sleep rhythm is, for example, the social jet lag (or this social jet It may be data that identifies a value obtained by calculating the lag using an arbitrary method. Here, for example, if a sample user's weekday sleep time is from 12:00 a.m. to 6:00 a.m., the central time is 3:00 a.m., and the sample user's holiday sleep time is from 12:00 a.m. to 6:00 a.m. If it is 3:00 AM to 11:00 AM, the social jet lag is 4 hours because the central time is 7:00 AM. Generally, it can be said that the smaller (or larger) the social jet lag, the better (or worse) the sleep rhythm of the sample user.

In one example, the sample user's sleep time periods on weekdays and holidays can be extracted from the sample user's answers to questions written in the Munich Chronotype Questionnaire (MCTQ) (see FIGS. 8A and 8B). However, in another example, it may be obtained from the sample user via email, web page, or the like.

Third, the sample time data may be data that identifies the sample user's sleep time (sleep time per day). In one example, the sample user's sleep time may be extracted from the sample user's responses to questions listed in the Munich Chronotype Questionnaire (MCTQ) (see FIGS. 8A and 8B), and in another example, It can be obtained from the sample user via email, web page, etc.

Fourth, the sample sleep category data may be data that identifies the sleep category to which (the sleep of) the sample user belongs. As an example, the sample sleep category data may be data that identifies the category to which (the sleep of) the sample user belongs, determined based on the sample property data of the sample user and the sample rhythm data of the sample user. . In this case, the sleep category may be determined, for example, by the following method.

FIG. 9 is a diagram showing an example of sleep category data used in the communication system 1 shown in FIG. 1. As shown in Figure 9, the score identified by the sample property data is arranged on one of the vertical and horizontal axes (here, the horizontal axis), and the score identified by the sample property data is arranged on the other of the vertical and horizontal axes (here, the vertical axis). Values identified by rhythm data may be placed. On one of the above-mentioned axes, at least one (here, one) threshold value, that is, threshold value A1 (6 points) may be set for the score identified by the sample property data. On the other axis, at least one (here, one) threshold value, that is, threshold value B1 (one hour) may be set for the value identified by the sample rhythm data. Thereby, four mutually different sleep categories (here, sleep categories 10, 11, 20, and 21) can be formed by the threshold A1 and the threshold B1, which are divided from the viewpoint of sleep characteristics and sleep rhythm. In this example, sleep category 10 is the category in which the disorder occurring in sleep is most likely to be significant, and sleep category 21 is the category in which the disorder occurring in sleep is most likely to be the least significant, It can be said that sleep categories 11 and 20 are categories positioned between sleep categories 10 and 21. Sleep category data may be data that identifies any sleep category among these four sleep categories.

FIG. 10 is a diagram showing another example of sleep category data used in the communication system 1 shown in FIG. 1. In the example shown in FIG. 10, two thresholds, for example, threshold A1 (6 points) and threshold A2 (8 points) may be set on one axis for the score identified by the sample property data. . On the other axis, two thresholds, for example, threshold B1 (1 hour) and threshold B2 (3 hours) may be set for the values identified by the sample rhythm data. As a result, nine mutually different sleep categories (here, sleep categories 100, 101, 102, 200, 201 , 202, 300, 301, 302) may be formed. In this example, the sleep category 300 is the category in which the disorder occurring in sleep is most likely to be significant, and the sleep category 202 is the category in which the disorder occurring in sleep is most likely to be the least significant, The remaining sleep categories can be said to be categories positioned between sleep categories 300 and 202. The sleep category data may be data that identifies any sleep category among these nine sleep categories.

The examples shown in FIGS. 9 and 10 are merely examples shown for convenience, and an arbitrary number of threshold values are set for the scores identified by the sample property data, and the values identified by the sample rhythm data. Any number of thresholds may be set for . By setting more thresholds for each of the sample property data and the sample rhythm data, it is possible to generate more sleep categories that are different from each other and are divided in terms of sleep properties and sleep rhythms. It is possible.

An information processing device, for example, a server device 10A (and/or a terminal device 20) managed by a company that operates an estimation service etc., stores such sample behavior data and sample sleep state data for each sample user. In combination, a set of training data can be generated and stored for the sample user. In order to achieve this, for example, the information processing device extracts sample behavior data attached with identification data that identifies a certain sample user from among the acquired plurality of sample behavior data, and Extract sample sleep state data attached with identification data that identifies the same sample user from the sleep state data, and combine them to generate and save a set of training data for this sample user. be able to. By performing similar processing for each of a plurality of sample users, the information processing device can generate and store multiple sets of teacher data.

(2) Step 1002
Returning to FIG. 3, next, in ST1002, the server device 10A (and/or the terminal device 20) managed by a company that operates an estimation service etc., which is an information processing device, An estimated model can be generated by training a learning model by machine learning using the training data of . Machine learning herein may include any well-known machine learning, such as, but not limited to, gradient boosting trees (eg, LightGBM), neural networks, or linear regression.

The information processing device can perform learning by sequentially inputting a plurality of sets of teacher data to a learning model. Here, as described above, each set of teacher data can include sample behavior data of the sample user corresponding to the set and sample sleep state data of the sample user corresponding to the set. The information processing device can use sample behavior data included in each set of teacher data as an explanatory variable, and use sample sleep state data included in that set of teacher data as an objective variable.

(2A) Example of machine learning For example, in a case where a gradient boosting tree is used, when sample behavior data (explanatory variables) included in a certain set of teacher data is input into learning data, the information processing device Decision trees can be created and added to the learning model to improve the objective function calculated from the data (objective variable) and estimated values (ie, the values output from the learning model). The information processing device can repeat this process for the number of decision trees determined by the hyperparameter. For the second and subsequent decision trees, it is possible to learn the difference between the objective variable and the estimated value from the decision trees created up to that point. In this way, the information processing device can determine the weights of branches and leaves of each decision tree. That is, in the gradient boosting tree, each decision tree uses explanatory variables of the teacher data as input data, and uses the values of leaf nodes reached by tracing each branch of the decision tree as output data (objective variables). Furthermore, the method of each branch of the decision tree and the value of each leaf node can be optimized so that the error between this data (target variable) and the target variable of the teacher data is small.

For example, in a case where the information processing device uses a neural network including an input layer, a middle layer, and an output layer, the information processing device inputs explanatory variables of the teaching data to the input layer, and outputs the explanatory variables from the output layer. It is possible to optimize (learn) each parameter (weight) included in the neural network using the error backpropagation method so as to reduce the error between the data (objective variable) and the objective variable in the training data. can.

In the case of using linear regression, the information processing device can learn the parameters of the linear regression model so as to improve the error between the target variable and the estimated value. This linear regression model is divided into an input layer and an output layer. Data (objective variables) output from the output layer by inputting explanatory variables from the teaching data into the input layer as input data, and data (objective variables) from the teaching data. Parameters included in the linear regression model can be optimized using the least squares method so that the error with the objective variable is small.

(2B) In an example of the first plurality of sets of teacher data included in the plurality of sets of teacher data , the information processing device calculates the estimation model using the first plurality of sets of teacher data included in the plurality of sets of teacher data. Can be generated (updated). Each set of teacher data included in the first plurality of sets of teacher data may include sample behavior data of a sample user corresponding to the group, and sample sleep state data of a sample user corresponding to the group. . Here, the sample sleep state data in each set of teacher data can include sample property data of the sample user corresponding to the set and sample rhythm data of the sample user.

In this case, the information processing device can perform preprocessing on each set of teacher data included in the first plurality of sets of teacher data. Specifically, the information processing device uses the sample property data and sample rhythm data included in each set of teacher data to perform the processing described above with reference to FIGS. 9 and 10, thereby completing the set. Preprocessing can be performed to generate sample sleep category data. As a result, the information processing device can process the preprocessed first plurality of sets of teacher data (each set of teacher data includes the sample behavior data of the sample user corresponding to the set and the sample sleep category data of the sample user). , etc.) can be input to the learning model to generate (update) the estimated model. As described below, such an estimation model is designed to identify the sleep category to which the target user belongs in response to inputting target behavior data that identifies the history of actions performed by the target user as an explanatory variable. At least sleep category data can now be output.

(2C) In an example of the second plurality of sets of teacher data included in the plurality of sets of teacher data , the information processing device calculates the estimation model using the second plurality of sets of teacher data included in the plurality of sets of teacher data. Can be generated (updated). Each set of teacher data included in the second plurality of sets of teacher data can include sample behavior data of a sample user corresponding to the group, and sample sleep state data of a sample user corresponding to the group. . Here, the sample sleep state data in each set of teacher data can include sample sleep category data of the sample user corresponding to the set.

In this case, the information processing device stores the second plurality of sets of teacher data (each set of teacher data includes sample behavior data of a sample user corresponding to the set and sample sleep category data of the sample user). can be input to the learning model to generate (update) the estimated model. As described below, such an estimation model is designed to identify the sleep category to which the target user belongs in response to inputting target behavior data that identifies the history of actions performed by the target user as an explanatory variable. At least sleep category data can now be output.

(2D) In an example of the third plurality of sets of teacher data included in the plurality of sets of teacher data , the information processing device calculates the estimation model using the third plurality of sets of teacher data included in the plurality of sets of teacher data. Can be generated (updated). Each set of teacher data included in the third plurality of sets of teacher data may include sample behavior data of a sample user corresponding to the set, and sample sleep state data of a sample user corresponding to the set. . Here, the sample sleep state data in each set of teacher data can include sample property data of the sample user corresponding to the set and sample rhythm data of the sample user.

In this case, the information processing device stores the third plurality of sets of teacher data (each set of teacher data includes the sample behavior data of the sample user corresponding to the set, the sample property data of the sample user, and the sample user's sample behavior data of the sample user corresponding to the set). sample rhythm data) can be input to the learning model to generate (update) the estimated model. As described below, such an estimation model is designed to identify the sleep characteristics of a target user in response to inputting target behavior data that identifies the history of actions performed by the target user as an explanatory variable. It becomes possible to output at least the property data and the target rhythm data that identifies the sleep rhythm of this target user.

(2E) In an example of the fourth plurality of sets of teacher data included in the plurality of sets of teacher data , the information processing device calculates the estimation model using the fourth plurality of sets of teacher data included in the plurality of sets of teacher data. Can be generated (updated). Each set of teacher data included in the fourth plurality of sets of teacher data may include sample behavior data of a sample user corresponding to the group, and sample sleep state data of a sample user corresponding to the group. . Here, the sample sleep state data in each set of teacher data may include sample property data of the sample user corresponding to the set.

In this case, the information processing device stores the fourth plurality of sets of teacher data (each set of teacher data includes the sample behavior data of the sample user corresponding to the set and the sample property data of the sample user). , can be input to a learning model to generate (update) an estimated model. As described below, such an estimation model is designed to identify the sleep characteristics of a target user in response to inputting target behavior data that identifies the history of actions performed by the target user as an explanatory variable. It becomes possible to at least output property data.

(2F) In an example of the fifth plurality of sets of teacher data included in the plurality of sets of teacher data , the information processing device calculates the estimation model using the fifth plurality of sets of teacher data included in the plurality of sets of teacher data. Can be generated (updated). Each set of teacher data included in the fifth plurality of sets of teacher data may include sample behavior data of a sample user corresponding to the group, and sample sleep state data of a sample user corresponding to the group. . Here, the sample sleep state data in each set of teacher data can include sample rhythm data of the sample user corresponding to that set.

In this case, the information processing device stores the fifth plurality of sets of teacher data (each set of teacher data includes the sample behavior data of the sample user corresponding to the set and the sample rhythm data of the sample user). , can be input to a learning model to generate (update) an estimated model. As described below, such an estimation model is designed to identify the sleep rhythm of a target user in response to inputting target behavior data that identifies the history of actions performed by the target user as an explanatory variable. At least rhythm data can now be output.

(2G) In an example of the sixth plurality of sets of teacher data included in the plurality of sets of teacher data , the information processing device calculates the estimation model using the fifth plurality of sets of teacher data included in the plurality of sets of teacher data. Can be generated (updated). Each set of teacher data included in the sixth plurality of sets of teacher data may include sample behavior data of a sample user corresponding to the set, and sample sleep state data of a sample user corresponding to the set. . Here, the sample sleep state data in each set of teacher data can include sample time data of a sample user corresponding to the set.

In this case, the information processing device stores the sixth plurality of sets of teacher data (each set of teacher data includes the sample behavior data of the sample user corresponding to the set and the sample time data of the sample user). , can be input to a learning model to generate (update) an estimated model. As described later, in response to inputting target behavior data that identifies the history of actions performed by the target user as an explanatory variable, such an estimation model identifies the target user's sleeping time. At least data can now be output.

The generation (update) of such an estimation model can be performed in place of the server device 10A (and/or terminal device 20) managed by a company that operates an estimation service, etc., or in place of the server device 10A (and/or terminal device 20) managed by a company that operates an estimation service, etc. The server device 10A (and/or terminal device 20) managed by another company can be executed by another server device 10 (and/or terminal device 20). The estimation model generated (updated) by the other server device 10 (terminal device 20) can be transmitted to any server device 10 including the server device 10A, or alternatively, the estimation model generated (updated) by the other server device 10 (terminal device 20) It can be accessed and used by the server device 10.

In addition, the names "first" to "sixth" added immediately before "multiple sets of teaching data" may be referred to as "multiple sets of teaching data" with a certain name among these names. , and ``multiple sets of training data'' with different names among these names, and are used for convenience, and are not intended to limit the order or content in any way.

(3) Step 1004
Referring again to FIG. 3, next, in ST1004, the server device 10A (and/or the terminal device 20), which is an information processing device and is managed by a company that operates an estimation service, etc., executes a specific application, for example. From the terminal device 20 (or other server device 10) that has accessed this server device 10A by executing (an application etc. for receiving provision of estimation services etc.) and/or by executing a browser and accessing this server device 10A. From the terminal device 20 that has accessed 10A, it is possible to acquire target action data that identifies the history of actions performed by the target user who is the estimation target. This target behavior data may be data that identifies at least one (one or more) behavior performed by the target user.

Such target behavior data can be acquired in place of the server device 10A (and/or terminal device 20) managed by a company that operates an estimation service, etc., or by using a server managed by a company that operates an estimation service, etc. Together with the device 10A (and/or the terminal device 20), it can be executed by another server device 10 (and/or the terminal device 20) managed by another company.

(4) Step 1006
Next, in ST1006, the information processing device, for example, the server device 10A (and/or the terminal device 20) managed by a company that operates an estimation service etc., uses the target behavior data acquired in ST1004 and Using the estimation model generated in , it is possible to obtain target sleep state data that identifies the sleep state of the target user. Specifically, the server device 10A can output target sleep state data as a target variable from the estimation model by inputting target behavior data as an explanatory variable into the estimation model.

The target sleep state data can include at least one of the following data.
・Data that identifies the sleep characteristics of the target user (“target characteristic data”)
・Data that identifies the target user's sleep rhythm ("target rhythm data")
・Data that identifies the target user's sleeping time ("target time data")
・Data that identifies the sleep category to which the target user (sleep) belongs (“target sleep category data”)

Which data the estimation model outputs as the target sleep state data may depend on which data is used as the sample sleep state data when the estimation model is generated (updated).

For example, when generating (updating) the estimation model, if the first plurality of sets of teacher data are used as sample sleep state data (paragraph 4(2)(2B) above), the estimation model can output at least target sleep category data as target sleep state data.

For example, when the estimation model is generated (updated), if the second plurality of sets of teacher data are used as sample sleep state data (paragraph 4(2)(2C) above), the estimation model can output at least target sleep category data as target sleep state data.

For example, when generating (updating) the estimation model, if the third set of teacher data is used as sample sleep state data (section 4(2)(2D) above), the estimation model can output at least target property data and target rhythm data as target sleep state data.

For example, when the fourth set of teacher data is used as sample sleep state data when generating (updating) the estimation model (paragraph 4(2)(2E) above), the estimation model can output at least target property data as target sleep state data.

For example, when generating (updating) the estimation model, if the fifth set of teacher data is used as sample sleep state data (paragraph 4 (2) (2F) above), the estimation model can output at least target rhythm data as target sleep state data.

For example, when generating (updating) the estimation model, if the sixth set of teacher data is used as sample sleep state data (paragraph 4(2)(2G) above), the estimation model can output at least target time data as target sleep state data.

For example, an estimation model is generated (updated) using the first plurality of sets of teacher data as sample sleep state data (section 4(2)(2B) above), and then the sample sleep state data When the estimation model is generated (updated) using the sixth set of teacher data (section 4(2)(2G) above), the estimation model is generated using the sixth set of teacher data. , target sleep category data and target time data can be output at least. Furthermore, if the estimation model is subsequently generated (updated) using the fifth set of training data as sample sleep state data (paragraph 4 (2) (2F) above), the estimation The model can output at least target sleep category data, target time data, and target rhythm data as target sleep state data.

As a method of using the estimation model, it is possible to use any of the methods illustrated below.
- The server device 10A acquires and holds the estimated model generated in ST1004, inputs target behavior data into this estimated model, and acquires target sleep state data output from this estimated model.
- The server device 10A transmits the target behavior data via the communication network 2 to an external device (any other server device 10 or any terminal device 20) that acquires and holds the estimation model generated in ST1004. By transmitting this, the estimation model that inputs this target behavior data as an explanatory variable is caused to output target sleep state data. Thereafter, the server device 10A receives the target sleep state data output by this estimation model from the external device via the communication network 2.

Such acquisition of target sleep state data is performed instead of the server device 10A (and/or terminal device 20) managed by a company that operates an estimation service, etc., or managed by a company that operates an estimation service, etc. Together with the server device 10A (and/or the terminal device 20), it can be executed by another server device 10 (and/or the terminal device 20) managed by another company.

(5) Step 1008
Next, in ST1008, the information processing device, for example, the server device 10A (and/or the terminal device 20) managed by a company that operates an estimation service etc. uses the target sleep state data acquired in ST1006 to At least one target product/one target service corresponding to this target sleep state data can be determined. Such target products/services may be suitable for the current (estimated) sleep state of the target user identified by the target sleep state data.

Specifically, in one example, the server device 10A can first prepare and store a search table that stores target sleep state data in association with a plurality of products and/or a plurality of services. Next, by inputting the target sleep state data acquired in ST1006 into the search table, the server device 10A inputs the data identifying at least one target product and/or at least one target service from this search table (“proposed data") can be obtained.

FIG. 11 is a diagram conceptually showing an example of a search table used in the communication system 1 shown in FIG. 1. The search table illustrated in FIG. 11 includes at least one item to be proposed in association with each data (target property data, target rhythm data, target time data, and target sleep category data) that may be included in the target sleep state data. At least one target product and/or at least one target service are stored.

In this example, in order to simplify the explanation, at least one of the three levels to be proposed is shown in association with each of the three mutually different levels indicated by each of the target property data, target rhythm data, target time data, and target sleep category data. At least one eligible product and/or at least one eligible service may be assigned. Taking the target property data as an example, the target property data indicates class V1 (a class in which the target user's sleep quality is good), class V2 (a class in which the target user's sleep quality is standard), and class V3 (a class in which the target user's sleep quality is standard). At least one target product and/or at least one target service to be proposed may be assigned in association with each class (class in which the target user has poor sleep characteristics).

Classes W1 to W3 may also indicate, in this order, whether the target user's sleep rhythm is "good," "normal," or "poor." Classes X1 to X3 may also indicate, in this order, whether the target user's sleep time is "long," "normal," or "short."

Classes Y1 to Y3 may also indicate, in this order, the sleep categories to which (the sleep of) the target user belongs: "good," "standard," and "poor." In one example, class Y1 corresponds to "sleep category 21" illustrated in FIG. 9, class Y2 corresponds to "sleep category 20" or "sleep category 11" in the diagram, and class Y3 corresponds to "sleep category 21" in the diagram. It can also be considered that it corresponds to "Sleep Category 10".

Products P1 to P11 illustrated in FIG. 11 all represent arbitrary products suitable for improving the user's sleeping state, and services S1 to S8 illustrated in FIG. Indicates any service suitable for improving sleep conditions. These products can include, but are not limited to, foods, daily necessities, medicines, athletic equipment, electrical products, and the like. Additionally, these services may include, but are not limited to, hospitals, clinics, gyms, and the like.

For example, when the target property data in the target sleep state data indicates class V3, the server device 10A can determine the product P4 and the service S1 as the target product and service, respectively. In addition, for example, when the target sleep state data includes target time data (indicating class , S4 can be determined as the target service.

Note that for each data included in the target sleep state data, the total number of classes used, the total number of products and/or services assigned to each class, etc. can be arbitrarily determined.

In another example, the server device 10A determines at least one target product and/or at least one target service using another estimation model generated by performing supervised learning (machine learning). is also possible. The other estimation model uses data indicating each class of each data included in the target sleep state data as an explanatory variable, and data indicating at least one target product and/or at least one target service as an objective variable, It can be generated (updated) by performing supervised learning.

The machine learning for learning the other estimation model may be, for example, any well-known machine learning, such as gradient boosting tree (e.g., LightGBM), neural network, linear regression, etc., as described in section "2A" above. may include, but are not limited to,

By inputting the target sleep state data into the other estimation model, the server device 10A generates data ("proposed data") that identifies at least one target product and/or at least one target service from the other estimation model. ) can be output.

As a method of using the other estimation model, it is possible to use any of the methods exemplified below.
- The server device 10A acquires and holds the generated another estimation model, inputs the target sleep state data into the other estimation model, and acquires the proposed data output from the other estimation model. .
- The server device 10A transmits the target sleep information via the communication network 2 to an external device (any other server device 10 or any terminal device 20) that acquires and holds the generated another estimation model. By transmitting the state data, the other estimation model that has input this target sleep state data as an explanatory variable is caused to output proposal data. Thereafter, the server device 10A receives the proposal data output by the other estimation model from the external device via the communication network 2.

Acquisition of such proposal data may be performed in place of the server device 10A (and/or terminal device 20) managed by a company that operates an estimation service, etc., or by using a server device managed by a company that operates an estimation service, etc. 10A (and/or the terminal device 20) as well as other server devices 10 (and/or terminal devices 20) managed by other companies.

(6) Step 1010
Referring again to FIG. 3, next, the information processing device, for example, the server device 10A (and/or the terminal device 20) managed by a company that operates an estimation service etc., uses the proposal data acquired in ST1008. can be provided (transmitted) via the communication network 2 to at least one terminal device 20 (which may be another server device 10) managed by a corporate user and/or an individual user. The "individual user" here may include the "target user."

Proposal data can be provided to corporate users and/or individual users by any means including web pages, e-mails, chats, electronic files, recording media (USB memory, CD-ROM, etc.), and/or paper media, etc. may be provided to at least one terminal device 20 (which may be another server device 10) managed by the server.

Such proposal data may be provided in place of the server device 10A (and/or terminal device 20) managed by a company that operates an estimation service, etc., or by a server device managed by a company that operates an estimation service, etc. 10A (and/or the terminal device 20) as well as other server devices 10 (and/or terminal devices 20) managed by other companies.

5. Modified Examples In the various examples described above, cases have been described in which an estimation model is generated and utilized using sample property data, sample rhythm data, sample time data, and sample sleep category data as objective variables. In another example, at least one (up to three) of sample property data, sample rhythm data, sample time data, and sample sleep category data are used as explanatory variables rather than objective variables to generate and use the estimation model. It is also possible to do so.

Furthermore, in the various examples described above, the execution entity of the various steps (for example, ST1000 to ST1010) illustrated in FIG. 3 is mainly the server device 10A (and/or The case where the terminal device 20) is used has been explained. However, it is also possible to have arbitrary plurality of server devices 10 and/or plurality of terminal devices 20 share and execute the various steps illustrated in FIG. 3 . Here, the arbitrary plurality of server devices 10 and/or plurality of terminal devices 20 may include, for example, a server device 10A (and/or terminal device 20) managed by a company that operates an estimation service etc. In another example, the server device 10A (and/or the terminal device 20) managed by a company that operates an estimation service or the like is not included.

In addition, in the above "4(4) step 1006", the definitions of each of the target property data, target rhythm data, target time data, and target sleep category data that may be included in the target sleep state data are shown. In another embodiment, the target sleep state data can include at least one of the following data instead of the data shown in "4(4) Step 1006" above.
・Data that “uniquely” identifies the sleep characteristics of the target user (“target characteristic data”)
・Data that "uniquely" identifies the target user's sleep rhythm ("target rhythm data")
・Data that “uniquely” identifies the sleep time of the target user (“target time data”)
・Data that “uniquely” identifies the sleep category to which the target user (sleep) belongs (“target sleep category data”)

Here, "target property data" "uniquely identifies" means that the sleep property of the target user can be identified without using any other data other than this "target property data". means. In other words, in a case where multiple pieces of data work together to identify the sleep characteristics of the target user, these multiple pieces of data do not fall under the category of data that "uniquely identifies" the target user's sleep characteristics. . For example, the first data includes data indicating that the target user's sleep quality is "low quality" and a flag "0", and the second data includes data indicating that the target user's sleep quality is "standard". Consider a case where the third data includes data indicating that the sleep quality of the target user is "high quality" and a flag "1". In this case, the sleep characteristics of the target user cannot be determined using only the first data and the second data. Only by checking the third data having the flag "1" can it be determined that the target user's sleep quality is "high quality." Therefore, none of these data is data that "uniquely" identifies the target user's sleep quality.

The term "uniquely identified" included in each of the above definitions of "target rhythm data," "target time data," and "target sleep category data" can be considered in the same way.

As explained above, according to the technology disclosed in the present application, by inputting target behavior data that identifies the history of at least one behavior performed by a target user into an estimation model, Target sleep state data may be obtained that identifies the user's current (predicted) sleep state. Furthermore, from among the plurality of target products and/or the plurality of target services prepared in advance, at least one target product and/or at least one target service suitable for the target sleep state data acquired in this way is provided to the target user. can be proposed to any user, including the target user, in order to improve the sleep state of the user.

Thereby, there is provided a computer program, an information processing apparatus, and an information processing apparatus that are used to estimate a sleep state of a target user based on a history of at least one action performed by the target user, and that have at least partially improved performance. method can be provided.

6. Operations performed by the communication system 2 to provide menopausal status data regarding the status of menopausal symptoms Next, specifics of the operations performed by the communication system 2 described above to provide menopausal status data regarding the status of menopausal symptoms. An example will be further described with reference to FIG. FIG. 12 is a flow diagram showing another example of operations performed by the communication system 1 shown in FIG. 1.

(1) Step 1100
First, in ST1100, an information processing device, for example, a server device 10A managed by a company that operates an estimation service etc., acquires and stores multiple sets of teacher data used to generate an estimation model. Can be done.

Each set of teacher data that constitutes a plurality of sets of teacher data includes data that identifies the history of actions performed by one sample user corresponding to this set ("sample action data"), and data that identifies the history of actions performed by one sample user corresponding to this set. Data identifying the state of menopausal symptoms of one sample user ("sample menopausal state data") can be included. For example, the first set of teacher data includes sample behavior data that identifies the history of actions performed by one sample user (Mr. V) corresponding to the first set, and the state of menopausal symptoms of Mr. V. Sample menopausal state data, and the second set of teacher data includes sample behavior data that identifies a history of actions performed by one sample user (Mr. W) corresponding to the second set; sample menopausal status data that identifies the status of Ms. W's menopausal symptoms.

(1A) Sample Behavior Data Each sample behavior data may be data that identifies at least one (one or more) behaviors out of a plurality of predetermined behaviors. Here, each of the plurality of predetermined actions may be an action that can at least partially have a positive or negative impact on the state of menopausal symptoms of the person who performed the action. Such a plurality of predetermined actions can include, for example, without being limited to, those exemplified below.
(Examples of actions that can at least partially positively influence the state of menopausal symptoms in humans)
・Performed aerobic exercise for at least 1 hour at least 2 days a week.
-Woke up at a certain time every day.
・I set aside time every week to relax.
Purchased 10 or more cans of beer (this behavior can be identified, for example, by data including the product category to which the beer belongs and its purchase quantity).
(Examples of actions that can at least partially adversely affect the state of menopausal symptoms in humans)
・Performed aerobic exercise for 1 hour or more no more than once a week.
-Woke up at a different time every day.
・I worked without taking a day off a week.
Purchased 5 or more bottles of green juice (this behavior can be identified, for example, by data including the product category to which the green juice belongs and the quantity purchased).

Each sample behavior data may be data that identifies at least one behavior among the plurality of predetermined behaviors illustrated in this way.

In addition, in one example, since the products purchased by the user and/or the services used by the user may tend to have a positive or negative impact on the state of menopausal symptoms of the user, the above-mentioned predetermined actions may be performed. can also include products purchased by the sample user (or the category to which the product belongs) and services used by the sample user (or the category to which the service belongs). This type of behavior can be generated using, for example, POS (Point of Sales) data held by retail stores (drug stores, supermarkets, convenience stores, etc.).

In one example, each of the plurality of predetermined behaviors is assigned unique identification data (such as letters, numbers, or a combination thereof), and each sample behavior data may include such identification data. can. For example, when using POS data, each of the plurality of sample behavior data includes JICFS classification (classification code), product category, purchase amount, purchase quantity, number of purchases, and/or corresponding to the purchased product. , date of purchase, etc., without limitation.

The server device 10A, which is an information processing device, for example, managed by a company that operates an estimation service, etc., can acquire such sample behavior data, for example, by at least one method among the methods exemplified below. can.
- The information processing device can access at least one other server device 10 (for example, another server device 10 installed in a drug store etc. that stores POS data, or this POS data) via the communication network 2. POS data is received from another server device 10).
- The information processing device sends, via the communication network 2, an email or a web page containing answers to a predetermined questionnaire (questionnaire regarding the behavior performed by the sample user) to at least one terminal device 20 and/or at least one Received from server device 10.
- The information processing device receives the contents of the responses via the communication network 2 from at least one server device 10 that has collected responses from a plurality of users to the above-mentioned predetermined questionnaire.
- The information processing device receives the contents of the responses to the above-mentioned predetermined questionnaire via a recording medium (USB memory, DVD-ROM, etc.) that has collected responses from a plurality of users.

(1B) Sample menopausal state data The sample menopausal state data included in each set of teacher data can include at least one of the data exemplified below, without being limited thereto.
・Data that identifies the overall status of one sample user corresponding to the group ("sample overall status data")
・Data identifying the state of the first symptom of one sample user corresponding to the group ("sample first state data")
・Data identifying the state of the second symptom of one sample user corresponding to the group ("sample second state data")

First of all, the sample overall condition data may be data that identifies whether the state of menopausal symptoms of the sample user is good (for example, how good it is) or bad (for example, how bad it is). In one example, the sample comprehensive state data may be data that identifies a score that comprehensively evaluates the state of menopausal symptoms of the sample user. In this case, the state of menopausal symptoms of the sample user may be, for example, data identifying a score calculated by any of the methods exemplified below.

・Based on the sample user's answers to the questions listed in the Simplified Menopause Index (SMI) (see Figure 13), the score is calculated using the method described in the SMI, or this score is calculated using an arbitrary method. Modified score obtained by (addition, subtraction, multiplication, division, etc.) ・Calculated using the method described in the Menopause Rating Scale based on the sample user's answers to the questions listed in the Menopause Rating Scale or a modified score obtained by calculating this score using any method (addition, subtraction, multiplication, division, etc.) ・Based on the sample user's answers to the questions listed in the Kupperman index , a score calculated by the method described in the Kupperman Menopause Index, or a modified score obtained by calculating this score using any method (addition, subtraction, multiplication, division, etc.) - A score calculated using the method described in the Kupperman Menopause Index A score calculated by the method described in the menopausal disorder evaluation based on the sample user's answers to the questionnaire, or a modified score obtained by calculating this score using any method (addition, subtraction, multiplication, division, etc.)・Based on the sample user's answers to the questions described in PSST (Premenstrual Symptoms Screening Tool), the score is calculated using the method described in PSST, or this score is calculated using any method (addition, subtraction, multiplication, division, etc.) ) ・Score calculated by the method described in WHQ (Women's Health Questionnaire) based on the sample user's answers to the questions described in WHQ (Women's Health Questionnaire), or this score Corrected score obtained by calculating using any method (addition, subtraction, multiplication, division, etc.) ・Calculated using the method described in the VAS (Visual Analogue Scale) based on the sample user's answers to the questions described in the VAS or a modified score obtained by calculating this score using an arbitrary method (addition, subtraction, multiplication, division, etc.) - Based on the sample user's answers to the questions listed in HFRDI (Hot Flash Related Daily Interference Scale) Based on the score calculated using the method described in HFRDI, or a modified score obtained by calculating this score using any method (addition, subtraction, multiplication, division, etc.) A score calculated by the method described in HFCS based on the sample user's answers to the questionnaire, or a modified score obtained by calculating this score using any method (addition, subtraction, multiplication, division, etc.) ・MENQOL ( A score calculated using the method described in MENQOL based on the sample user's answers to the questions described in Menopause-Specific Quality of Life), or a calculation of this score using any method (addition, subtraction, multiplication, division, etc.) Corrected score obtained by ・Calculated using the method described in the Menopausal Symptom Evaluation Form for Japanese Women, based on the sample user's answers to the questions listed in the Japanese Women's Menopausal Symptom Evaluation Form. Score, or a modified score obtained by calculating this score using any method (addition, subtraction, multiplication, division, etc.) Based on the sample user's answers to the questions listed on any other questionnaire or scale, A score calculated using the method described in the questionnaire or scale, or a modified score obtained by calculating this score using any method.

Second, the sample first state data identifies whether the state of one symptom of the sample user among a plurality of menopausal symptoms is good (e.g., how good it is) or bad (e.g., how bad it is). It can be data. In one example, the sample first state data may be data that identifies a score that evaluates the state of the one symptom of the sample user.

Taking the SMI as an example, the above-mentioned multiple menopausal symptoms may be multiple menopausal symptoms described in the SMI such as "hot face," "prone to sweating," and "easiness to get cold in the lower back and limbs." In addition to SMI, the plurality of menopausal symptoms described above may be a plurality of menopausal symptoms listed in any of the indices listed above. Further, the above-mentioned one symptom may be any one symptom selected from such a plurality of menopausal symptoms.

Taking the SMI as an example, the sample first state data can be extracted from the answers given by the sample user to the questionnaires listed in the SMI (see FIG. 13), in one example. In this case, the sample first state data is data that identifies the state of one symptom, such as "shortness of breath and palpitations," and has a score of 0, 4, 8, or 12 (see FIG. 13). It can be identifying data. In another example, sample first state data may be obtained from the sample user via email, web page, or the like.

Third, the sample second state data identifies whether the state of another symptom of the sample user among the plurality of menopausal symptoms is good (e.g., how good is it) or bad (e.g., how bad is it)? It can be data. In one example, the sample second state data may be data that identifies a score that evaluates the state of the another symptom of the sample user.

Taking the SMI as an example, the above-mentioned multiple menopausal symptoms may be multiple menopausal symptoms described in the SMI such as "hot face," "prone to sweating," and "easiness to get cold in the lower back and limbs." In addition to SMI, the plurality of menopausal symptoms described above may be a plurality of menopausal symptoms listed in any of the indices listed above. Further, the above-mentioned another symptom may be any one symptom selected from such a plurality of menopausal symptoms, but may be a symptom different from the above-mentioned one symptom.

Taking the SMI as an example, the sample second status data can be extracted from the answers given by the sample user to the questionnaires listed in the SMI (see FIG. 13), in one example. In this case, the sample second state data is, for example, data that identifies another symptom state such as "easily fatigued", and for example, data that identifies a score of 0, 2, 4, or 7 (see FIG. 13). It can be. In another example, sample second state data may be obtained from the sample user via email, web page, or the like.

For example, the server device 10A (and/or terminal device 20), which is an information processing device managed by a company that operates an estimation service, etc., stores such sample behavior data and sample menopausal status data for each sample user. In combination, a set of training data can be generated and stored for the sample user. In order to achieve this, for example, the information processing device extracts sample behavior data attached with identification data that identifies a certain sample user from among the acquired plurality of sample behavior data, and Extract sample menopausal state data attached with identification data that identifies the same sample user from the menopausal state data, and combine them to generate and save a set of training data for this sample user. be able to. By performing similar processing for each of a plurality of sample users, the information processing device can generate and store multiple sets of teacher data.

(2) Step 1102
Returning to FIG. 3, next, in ST1102, the server device 10A (and/or the terminal device 20) managed by a company that operates an estimation service etc., which is an information processing device, An estimated model can be generated by training a learning model by machine learning using the training data of . Machine learning herein may include any well-known machine learning, such as, but not limited to, gradient boosting trees (eg, LightGBM), neural networks, or linear regression.

(2A) In an example of the first plurality of sets of teacher data included in the plurality of sets of teacher data , the information processing device calculates the estimation model using the first plurality of sets of teacher data included in the plurality of sets of teacher data. Can be generated (updated). Each set of teacher data included in the first plurality of sets of teacher data can include sample behavior data of a sample user corresponding to the group, and sample menopausal status data of a sample user corresponding to the group. . Here, the sample menopausal state data in each set of teacher data can include sample comprehensive state data of the sample user corresponding to that set.

In this case, the information processing device stores the first plurality of sets of teacher data (each set of teacher data includes sample behavior data of the sample user corresponding to the set and sample comprehensive state data of the sample user). can be input to the learning model to generate (update) the estimated model. As described below, such an estimation model identifies the overall state of the target user in response to inputting target behavior data that identifies the history of actions performed by the target user as an explanatory variable. It becomes possible to output at least the target comprehensive state data.

In one embodiment, the sample menopausal state data in each set of teacher data identifies the "score corresponding to" the overall state of menopausal symptoms of one sample user corresponding to the set. think of.

In this case, the information processing device can perform preprocessing on each set of teacher data included in the first plurality of sets of teacher data. Specifically, for each set of teacher data, the information processing device assigns the score identified by the sample comprehensive state data corresponding to this set to one of a plurality of categories using at least one threshold value. You can perform preprocessing to convert it into categories. For example, when using one threshold (first threshold), the information processing device determines whether the score identified by the sample comprehensive state data corresponding to each set is equal to or higher than the first threshold. It can be converted into one of two categories (“High” and “Low”) depending on whether it is less than or equal to the maximum value. Alternatively, for example, when using two thresholds (a first threshold and a second threshold smaller than the first threshold), the information processing device may calculate the score identified by the sample comprehensive state data corresponding to each set. , the score is divided into three categories ("High", "Middle" and "Low") depending on whether the score is greater than or equal to the first threshold, greater than or equal to the second threshold and less than the first threshold, or less than the second threshold. ”) can be converted into any of the following categories.

In this way, the information processing apparatus is configured to determine, for each set of teacher data included in the first plurality of sets of teacher data, the information processing apparatus according to Preprocessing is performed to convert sample comprehensive state data that identifies the "score" into sample comprehensive state data that identifies the "category corresponding to" the comprehensive state of menopausal symptoms of one sample user corresponding to this set. can be executed.

As a result, the information processing device can process the preprocessed first plurality of sets of teacher data (each set of teacher data includes the sample behavior data of the sample user corresponding to the set and the sample comprehensive state data of the sample user). , etc.) can be input to the learning model to generate (update) the estimated model.

As described below, such an estimation model calculates the overall state of menopausal symptoms of the target user in response to inputting target behavior data that identifies the history of actions performed by the target user as an explanatory variable. It becomes possible to output at least the object comprehensive state data that identifies the category corresponding to the category.

(2B) In an example of the second plurality of sets of teacher data included in the plurality of sets of teacher data , the information processing device calculates the estimation model using the second plurality of sets of teacher data included in the plurality of sets of teacher data. Can be generated (updated). Each set of teacher data included in the second plurality of sets of teacher data can include sample behavior data of a sample user corresponding to the set, and sample menopausal status data of a sample user corresponding to the set. . Here, the sample menopausal state data in each set of teacher data can include sample first state data of the sample user corresponding to that set.

In this case, the information processing device includes the second plurality of sets of teacher data (each set of teacher data includes sample behavior data of a sample user corresponding to the set and sample first state data of the sample user). ) can be input to the learning model to generate (update) the estimated model. As described below, such an estimation model identifies the state of the first symptom of the target user in response to inputting target behavior data that identifies the history of actions performed by the target user as an explanatory variable. At least the target first state data can be output.

In one embodiment, the sample first state data in each set of teacher data identifies the "score corresponding to" the state of the first symptom of one sample user corresponding to the set. think.

In this case, the information processing device can perform preprocessing on each set of teacher data included in the second plurality of sets of teacher data. Specifically, for each set of teacher data, the information processing device assigns the score identified by the sample first state data corresponding to this set to one of a plurality of categories using at least one threshold value. It is possible to perform preprocessing to convert it into a category. For example, when using one threshold value (first threshold value), the information processing device determines whether the score identified by the sample first state data corresponding to each set is equal to or higher than the first threshold value or Depending on whether it is less than the threshold, it can be converted into one of two categories (“High” and “Low”). Alternatively, for example, in the case of using two thresholds (a first threshold and a second threshold smaller than the first threshold), the information processing device may calculate the score identified by the sample first state data corresponding to each set. is divided into three categories ("High", "Middle" and " ``Low'').

In this way, the information processing device determines, for each set of teacher data included in the second plurality of sets of teacher data, a score corresponding to the state of the first symptom of one sample user corresponding to this set. Preprocessing is performed to convert the sample first state data that identifies the first symptom into the sample first state data that identifies the "category corresponding to" the state of the first symptom of one sample user corresponding to this set. be able to.

As a result, the information processing device obtains the preprocessed second plurality of sets of teacher data (each set of teacher data includes the sample behavior data of the sample user corresponding to the set and the sample first state data of the sample user). ) can be input to the learning model to generate (update) the estimated model.

As described below, such an estimation model responds to the state of the first symptom of the target user in response to inputting target behavior data that identifies the history of actions performed by the target user as an explanatory variable. It is now possible to output at least the target first state data that identifies the category of the target.

Furthermore, in addition to the above-described preprocessing ("preprocessing A" for convenience), the information processing apparatus can also perform further preprocessing ("preprocessing B" for convenience). Specifically, first, the information processing device can perform "preprocessing B" of extracting another plurality of sets of teacher data from the second plurality of sets of teacher data. Each set of teacher data included in the other plurality of sets of teacher data has sample first state data that identifies scores equal to or higher than the threshold.

For example, when the first symptom is "sweating easily" in the SMI (see Figure 13), each set of teacher data included in the other plural sets of teacher data has a score ( 10 or 6). That is, each set of teacher data included in the other plurality of sets of teacher data includes teacher data related only to the sample user who has the stronger first symptom (here, "prone to sweating").

Further, the information processing device can use the other plurality of sets of teacher data extracted in this way as a "new" second plurality of sets of teacher data. Thereafter, the information processing device can perform the above-described preprocessing A on the "new" second plurality of sets of teacher data. In this preprocessing A, at least one threshold value to be used can be set to a larger value than in the above-mentioned example in which preprocessing B is not performed. Next, the information processing apparatus can generate (update) the estimated model by inputting the "new" second plurality of sets of teacher data after the preprocessing A into the learning model.

Such "new" second plurality of sets of teacher data includes sample behavior data and sample first state data only for sample users who have stronger first symptoms. Therefore, an estimation model that is trained using such "new" second sets of training data is able to classify target users who have performed actions that may lead to strong first symptoms into strong first symptoms. It is possible to more easily estimate the level of intensity of the first symptom of such a target user, and it is also possible to more precisely estimate the level of the intensity of the first symptom of such a target user.

(2C) In an example of the third plurality of sets of teacher data included in the plurality of sets of teacher data , the information processing device calculates the estimation model using the third plurality of sets of teacher data included in the plurality of sets of teacher data. Can be generated (updated). Each set of teacher data included in the third plurality of sets of teacher data can include sample behavior data of a sample user corresponding to the group, and sample menopausal status data of a sample user corresponding to the group. . Here, the sample menopausal state data in each set of teacher data can include sample second state data of the sample user corresponding to that set.

In this case, the information processing device includes the third plurality of sets of teacher data (each set of teacher data includes sample behavior data of the sample user corresponding to the set and sample second state data of the sample user). ) can be input to the learning model to generate (update) the estimated model. As described below, such an estimation model identifies the state of the second symptom of the target user in response to inputting target behavior data that identifies the history of actions performed by the target user as an explanatory variable. At least the target second state data can be output.

In one embodiment, the sample second state data in each set of teacher data identifies the "score corresponding to" the state of the second symptom of one sample user corresponding to the set. think.

In this case, the information processing device can perform preprocessing on each set of teacher data included in the third plurality of sets of teacher data. Specifically, for each set of teacher data, the information processing device assigns the score identified by the sample second state data corresponding to this set to one of a plurality of categories using at least one threshold value. It is possible to perform preprocessing to convert it into a category. For example, when using one threshold value (first threshold value), the information processing device determines whether the score identified by the sample second state data corresponding to each set is equal to or higher than the first threshold value. Depending on whether it is less than the threshold, it can be converted into one of two categories (“High” and “Low”). Alternatively, for example, in the case of using two thresholds (a first threshold and a second threshold smaller than the first threshold), the information processing device may calculate the score identified by the sample second state data corresponding to each set. is divided into three categories ("High", "Middle" and " ``Low'').

In this way, the information processing apparatus determines, for each set of teacher data included in the third plurality of sets of teacher data, a score corresponding to the state of the second symptom of one sample user corresponding to this group. Preprocessing is performed to convert the sample second state data that identifies the ``category corresponding to'' of the second symptom state of one sample user corresponding to this set into the sample second state data that identifies the ``category corresponding to''. be able to.

As a result, the information processing device obtains the preprocessed third plurality of sets of teacher data (each set of teacher data includes the sample behavior data of the sample user corresponding to the set and the sample first state data of the sample user). ) can be input to the learning model to generate (update) the estimated model.

As described later, such an estimation model responds to the state of the second symptom of the target user in response to inputting target behavior data that identifies the history of actions performed by the target user as an explanatory variable. It becomes possible to output at least the target second state data that identifies the category of the object.

Furthermore, in addition to the above-described preprocessing ("preprocessing A" for convenience), the information processing apparatus can also perform further preprocessing ("preprocessing B" for convenience). Specifically, first, the information processing apparatus can perform "preprocessing B" of extracting another plurality of sets of teacher data from the third plurality of sets of teacher data. Each set of teacher data included in the other plurality of sets of teacher data has sample second state data that identifies scores equal to or higher than the threshold.

For example, when the second symptom is the symptom of "getting tired easily" in the SMI (see Figure 13), each set of teacher data included in the other plural sets of teacher data has a score of 3 or more (7 or 4) may be sample second state data identifying the second state data. That is, each set of teacher data included in the other plurality of sets of teacher data includes teacher data related only to the sample user who has the stronger second symptom (here, "easily tired").

Further, the information processing device can use the other plurality of sets of teacher data extracted in this way as a "new" third plurality of sets of teacher data. Thereafter, the information processing apparatus can perform the above-described preprocessing A on the "new" third set of teacher data. In this preprocessing A, at least one threshold value to be used can be set to a larger value than in the above-mentioned example in which preprocessing B is not performed. Next, the information processing apparatus can generate (update) the estimated model by inputting the "new" third plurality of sets of teacher data after the preprocessing A into the learning model.

Such a "new" third plurality of sets of teacher data includes sample behavior data and sample second state data only for sample users who have stronger second symptoms. Therefore, an estimation model that is trained using such "new" third sets of training data will be able to classify a target user who has performed an action that may lead to a strong second symptom into a strong second symptom. It is possible to more easily estimate the level of the intensity of the second symptom of such a target user as a user who has the following symptoms.

The generation (update) of the estimation model as described above is performed in place of the server device 10A (and/or terminal device 20) managed by the company that operates the estimation service, etc., or by the company that operates the estimation service, etc. The server device 10A (and/or the terminal device 20) managed by the server device 10A (and/or the terminal device 20) may be executed by another server device 10 (and/or the terminal device 20) managed by another company. The estimation model generated (updated) by the other server device 10 (terminal device 20) can be transmitted to any server device 10 including the server device 10A, or alternatively, the estimation model generated (updated) by the other server device 10 (terminal device 20) It can be accessed and used by the server device 10.

In addition, the names "first" to "third" added immediately before "multiple sets of teaching data" may be referred to as "multiple sets of teaching data" with a certain name among these names. , and ``multiple sets of training data'' with different names among these names, and are used for convenience, and are not intended to limit the order or content in any way.

(3) Step 1104
Referring again to FIG. 12, next, in ST1104, the server device 10A (and/or the terminal device 20), which is an information processing device, managed by a company that operates an estimation service, etc., executes a specific application, for example. From the terminal device 20 (or other server device 10) that has accessed this server device 10A by executing (an application etc. for receiving provision of estimation services etc.) and/or by executing a browser and accessing this server device 10A. From the terminal device 20 that has accessed 10A, it is possible to acquire target action data that identifies the history of actions performed by the target user who is the estimation target. This target behavior data may be data that identifies at least one (one or more) behavior performed by the target user.

Such target behavior data can be acquired in place of the server device 10A (and/or terminal device 20) managed by a company that operates an estimation service, etc., or by using a server managed by a company that operates an estimation service, etc. Together with the device 10A (and/or the terminal device 20), it can be executed by another server device 10 (and/or the terminal device 20) managed by another company.

(4) Step 1106
Next, in ST1106, the information processing device, for example, the server device 10A (and/or the terminal device 20) managed by a company that operates an estimation service etc., uses the target behavior data acquired in ST1104 and Using the estimation model generated in , it is possible to obtain target menopausal state data that identifies the state of menopausal symptoms of the target user. Specifically, the server device 10A can output target menopausal state data as a target variable from the estimation model by inputting target behavior data as an explanatory variable into the estimation model.

The target menopausal status data can include at least one of the following data.
・Data that identifies the overall condition of the target user's menopausal symptoms ("target comprehensive status data")
・Data that identifies the state of the first symptom of the target user (“Target first state data”)
・Data identifying the state of the second symptom of the target user (“Target second state data”)

Which data the estimation model outputs as the target menopausal state data may depend on which data is used as the sample menopausal state data when the estimation model is generated (updated).

For example, when the estimation model is generated (updated), if the first plurality of sets of teacher data are used as sample menopausal status data (paragraph 6 (2) (2A) above), the estimation model can output at least target comprehensive state data as target menopausal state data.

For example, when the estimation model is generated (updated), if the second plurality of sets of teacher data are used as sample menopausal status data (section "6(2)(2B)" above), the estimation model can output at least the target first state data as the target menopausal state data.

For example, when generating (updating) the estimation model, if the third set of training data is used as sample menopausal state data (section 6 (2) (2C) above), the estimation model can output at least the target second state data as the target menopausal state data.

For example, an estimation model is generated (updated) using the first plurality of sets of training data as sample menopausal state data (section 6 (2) (2A) above), and then the sample menopausal state data If the estimation model is generated (updated) using the second plurality of sets of training data (Section 6(2)(2B) above), the estimation model is , the object comprehensive state data and the object first state data can be output at least. Furthermore, if the estimation model is subsequently generated (updated) using the third set of training data as sample menopausal status data (paragraph 6 (2) (2C) above), the estimation The model can output at least target comprehensive state data, target first state data, and target second state data as target menopausal state data.

As a method of using the estimation model, it is possible to use any of the methods illustrated below.
- The server device 10A acquires and holds the estimated model generated in ST1104, inputs target behavior data into this estimated model, and acquires target menopausal state data output from this estimated model.
- The server device 10A transmits the target behavior data via the communication network 2 to an external device (any other server device 10 or any terminal device 20) that acquires and holds the estimation model generated in ST1104. By transmitting this, the estimation model into which this target behavior data is input as an explanatory variable is caused to output target menopausal state data. Thereafter, the server device 10A receives the target menopausal state data output by this estimation model from the external device via the communication network 2.

Such target menopausal state data can be acquired in place of the server device 10A (and/or terminal device 20) managed by a company that operates an estimation service, etc., or managed by a company that operates an estimation service, etc. Together with the server device 10A (and/or the terminal device 20), it can be executed by another server device 10 (and/or the terminal device 20) managed by another company.

(5) Step 1108
Next, in ST1108, the information processing device, for example, the server device 10A (and/or the terminal device 20) managed by a company that operates an estimation service etc. uses the target menopausal state data acquired in ST1106, At least one target product/one target service corresponding to this target menopausal state data can be determined. Such target products/services may be suitable for the current (estimated) state of menopausal symptoms of the target user identified by the target menopausal status data.

Specifically, in one example, the server device 10A can first prepare and store a search table that stores target menopausal state data in association with a plurality of products and/or a plurality of services. Next, the server device 10A inputs the target menopausal state data acquired in ST1106 into the search table, thereby obtaining data identifying at least one target product and/or at least one target service (“proposed data") can be obtained.

FIG. 14 is a diagram conceptually showing another example of the search table used in the communication system 1 shown in FIG. 1. The search table illustrated in FIG. 14 includes at least one item to be proposed in association with each data (each of the target comprehensive state data, the target first state data, and the target second state data) that may be included in the target menopausal state data. At least one target product and/or at least one target service are stored.

In this example, in order to simplify the explanation, at least one of the three levels to be proposed is shown in association with each of the three mutually different levels indicated by each of the target comprehensive state data, the target first state data, and the target second state data. At least one eligible product and/or at least one eligible service may be assigned. Taking the target comprehensive state data as an example, the target comprehensive state data indicates class V1 (class where the comprehensive state of menopausal symptoms of the target user is good) and class V2 (class where the comprehensive state of menopausal symptoms of the target user is standard). At least one target product and/or at least one target service to be proposed is assigned in association with each of class V3 (a class in which the overall state of menopausal symptoms of the target user is poor) obtain.

Classes W1 to W3 may also indicate, in this order, that the state of the first symptom of the target user is "good", "normal", or "bad". Classes X1 to X3 may also indicate, in this order, the state of the second symptom of the target user as "good", "normal", or "bad".

The products illustrated in FIG. 14 (products with reference signs consisting of "P" and numbers, such as P10 and P20) are any products suitable for improving the state of menopausal symptoms of the user. The services illustrated in FIG. 14 (services with reference signs consisting of "S" and numbers, such as S10 and S20) are all suitable for improving the state of menopausal symptoms of the user. Indicates any service. These products can include, but are not limited to, foods, daily necessities, medicines, athletic equipment, electrical products, and the like. Additionally, these services may include, but are not limited to, hospitals, clinics, gyms, and the like.

For example, when the target comprehensive state data in the target menopausal state data indicates class V3, the server device 10A can determine the product P40 and the service S10 as the target product and the target service, respectively. In addition, for example, when the target menopausal state data includes target second state data (indicating class X2) and target first state data (indicating class W3), server device 10A sets product P80 as the target product, Services S10, S30, and S40 can be determined as target services.

Note that for each data included in the target menopausal state data, the total number of classes used, the total number of products and/or services allocated to each class, etc. can be arbitrarily determined.

In another example, the server device 10A determines at least one target product and/or at least one target service using another estimation model generated by performing supervised learning (machine learning). is also possible. The other estimation model uses data indicating each class of each data included in the target menopausal state data as an explanatory variable, and data indicating at least one target product and/or at least one target service as an objective variable, It can be generated (updated) by performing supervised learning.

The machine learning for learning the other estimation model may be, for example, any well-known machine learning, such as a gradient boosting tree (e.g., LightGBM), a neural network, or a linear regression, as described above. It can be included without

By inputting target menopausal state data into the other estimation model, the server device 10A generates data ("proposed data") that identifies at least one target product and/or at least one target service from the other estimation model. ) can be output.

As a method of using the other estimation model, it is possible to use any of the methods exemplified below.
- The server device 10A acquires and holds the generated another estimation model, inputs the target menopausal state data into the other estimation model, and acquires the proposed data output from the other estimation model. .
- The server device 10A transmits the target menopause via the communication network 2 to an external device (any other server device 10 or any terminal device 20) that acquires and holds the generated another estimation model. By transmitting the state data, the other estimation model that has input this target menopausal state data as an explanatory variable is caused to output proposed data. Thereafter, the server device 10A receives the proposal data output by the other estimation model from the external device via the communication network 2.

Acquisition of such proposal data may be performed in place of the server device 10A (and/or terminal device 20) managed by a company that operates an estimation service, etc., or by using a server device managed by a company that operates an estimation service, etc. 10A (and/or the terminal device 20) as well as other server devices 10 (and/or terminal devices 20) managed by other companies.

(6) Step 1110
Referring again to FIG. 12, next, the information processing device, for example, the server device 10A (and/or the terminal device 20) managed by a company that operates an estimation service etc., uses the proposal data acquired in ST1108. , can be provided (transmitted) via the communication network 2 to at least one terminal device 20 (which may be another server device 10) managed by a corporate user and/or an individual user. The "individual user" here may include the "target user."

Proposal data can be provided to corporate users and/or individual users by any means including web pages, e-mails, chats, electronic files, recording media (USB memory, CD-ROM, etc.), and/or paper media, etc. may be provided to at least one terminal device 20 (which may be another server device 10) managed by the server.

Such proposal data may be provided in place of the server device 10A (and/or terminal device 20) managed by a company that operates an estimation service, etc., or by a server device managed by a company that operates an estimation service, etc. 10A (and/or the terminal device 20) as well as other server devices 10 (and/or terminal devices 20) managed by other companies.

7. Modifications In the various examples described above, cases have been described in which an estimation model is generated and utilized using sample comprehensive state data, sample first state data, and sample second state data as target variables. In another example, an estimation model is generated and used by using at least one (up to two) of sample comprehensive state data, sample first state data, and sample second state data as explanatory variables instead of objective variables. It is also possible to do so.

Furthermore, in the various examples described above, the execution entity of the various steps (for example, ST1100 to ST1110) illustrated in FIG. 12 is mainly the server device 10A (and/or The case where the terminal device 20) is used has been explained. However, it is also possible to have arbitrary plurality of server devices 10 and/or plurality of terminal devices 20 share and execute the various steps illustrated in FIG. 12 . Here, the arbitrary plurality of server devices 10 and/or plurality of terminal devices 20 may include, for example, a server device 10A (and/or terminal device 20) managed by a company that operates an estimation service etc. In another example, the server device 10A (and/or the terminal device 20) managed by a company that operates an estimation service or the like is not included.

As explained above, according to the technology disclosed in the present application, by inputting target behavior data that identifies the history of at least one behavior performed by a target user into an estimation model, Target menopausal status data that identifies a user's current (predicted) menopausal symptom status may be obtained. Furthermore, from among the plurality of target products and/or the plurality of target services prepared in advance, at least one target product and/or at least one target service suitable for the target menopausal status data obtained in this way is provided to the target user. It can be proposed to any user including the target user in order to improve menopausal symptoms.

Thereby, a computer program, an information processing device, which is used for estimating the state of menopausal symptoms of a target user based on the history of at least one action performed by the target user, and has at least partially improved performance. and a method can be provided.

8. The technology disclosed in this application regarding the estimation model for predicting various health conditions is not only useful for predicting the above-mentioned sleep state and menopausal symptom state, but also for estimating various health conditions as exemplified below. It is also applicable to predicting states.
・Exercise status ・Immune status ・Hydration status ・Nutritional status

(1) Regarding the state of exercise , an estimation model for predicting the state of exercise is used to predict conditions related to lifestyle-related diseases. In order to generate such an estimation model, the sample behavior data that is an explanatory variable may be at least one of a plurality of data exemplified below.
・Data that identifies the products (or the category to which the products belong) that the sample user purchased during a certain period of time (for example, one year, etc.) ・The data that identifies the services (or the category to which the services belong) that the sample user used during a certain period of time (for example, one year, etc.)・Data that identifies whether the sample user is obese or not ・Data that identifies the amount of change in the sample user's weight over a certain period of time (for example, from the age of 30 to the present) ・Data that identifies the amount of exercise of the sample user (e.g. number of steps, etc.)

Furthermore, the data identifying the symptom, which is the target variable, may be at least one of a plurality of data (which may be a measured value) exemplified below.
・Data that identifies the sample user's weight ・Data that identifies the sample user's BMI (Body Mass Index) ・Data that identifies the sample user's step count ・Data that identifies the sample user's smoking status ・Metabolic syndrome diagnosis of the sample user Data that identifies reference values

Note that the metabolic syndrome diagnostic reference value includes the sample user's visceral fat accumulation and/or waist circumference length, and at least two values among the plurality of values exemplified below.
・Threshold values established for the diagnosis of hyperglyceridemia and/or hypoHDL cholesterolemia for the sample user. ・Systolic and/or diastolic blood pressure of the sample user. ・For the diagnosis of fasting hyperglycemia for the sample user. threshold set in

The information processing device can generate an estimation model that predicts the state of movement by performing machine learning using such explanatory variables and objective variables. The operations performed by the communication system 2 in connection with the generation and utilization method of such an estimation model are described in the above "6" and "7" above in relation to the aspect of predicting the state of menopausal symptoms. The operation may be substantially the same as described. In the explanations given in Sections “6” and “7” above (with particular reference to FIGS. 12 and 14), the explanatory variables and objective variables used in the estimation model for predicting the state of menopausal symptoms are as follows: Executed by the communication system 2 in connection with the generation and usage of an estimation model for predicting the state of movement by replacing them with the explanatory variables and objective variables described in Section 8(1) of this article, respectively. Operation will be understood by those skilled in the art.

As explained above, according to the technology disclosed in the present application, by inputting target behavior data that identifies the history of at least one behavior performed by a target user into an estimation model, Target state data identifying the user's current (predicted) movement state may be obtained. Furthermore, from among the plurality of target products and/or the plurality of target services prepared in advance, at least one target product and/or at least one target service suitable for the target status data acquired in this way is provided to the target user. Suggestions can be made to any user, including the target user, to improve the state of exercise.

Thereby, a computer program, an information processing apparatus, and a computer program, which are used to estimate the state of movement of a target user based on a history of at least one action performed by the target user, and which have at least partially improved performance. method can be provided.

(2) Regarding the immune status , the estimation model for predicting the immune status is used to predict the severity of symptoms of a person whose body's immune system is weakened. In order to generate such an estimation model, the sample behavior data that is an explanatory variable may be at least one of a plurality of data exemplified below.
・Data that identifies the products (or the category to which the products belong) that the sample user purchased during a certain period of time (for example, one year, etc.) ・The data that identifies the services (or the category to which the services belong) that the sample user used during a certain period of time (for example, one year, etc.) data that identifies categories)

Furthermore, the data identifying the symptom, which is the target variable, may be at least one of a plurality of data (which may be a measured value) exemplified below.
・Data that identifies how easily the sample user catches a cold ・Data that identifies how often the sample user catches a cold ・Data that identifies the SIgA concentration of the sample user ・Data that identifies the allergic symptoms of the sample user ・Oral cavity of the sample user Data that identifies the state of the environment and/or immunity ・Data that identifies the state of sleep of the sample user ・Data that identifies the state of exercise of the sample user ・Data that identifies the state of stress of the sample user

Note that the data for identifying such symptoms may be the sample user's subjective symptoms obtained from a questionnaire or the like, or may be obtained from the sample user's reception history at a medical institution. Data identifying such symptoms may identify any of multiple (preferably three or more) categories.

The information processing device can generate an estimation model that predicts the immune status by performing machine learning using such explanatory variables and objective variables. The operations performed by the communication system 2 in connection with the generation and utilization method of such an estimation model are described in the above "6" and "7" above in relation to the aspect of predicting the state of menopausal symptoms. The operation may be substantially the same as described. In the explanations given in Sections “6” and “7” above (with particular reference to FIGS. 12 and 14), the explanatory variables and objective variables used in the estimation model for predicting the state of menopausal symptoms are as follows: Executed by the communication system 2 in connection with the generation and usage method of an estimation model for predicting the immune status by replacing them with the explanatory variables and objective variables described in Section 8 (2) of this article, respectively. Operation will be understood by those skilled in the art.

As explained above, according to the technology disclosed in the present application, by inputting target behavior data that identifies the history of at least one behavior performed by a target user into an estimation model, Subject status data may be obtained that identifies the user's current (predicted) immune status. Furthermore, from among the plurality of target products and/or the plurality of target services prepared in advance, at least one target product and/or at least one target service suitable for the target status data acquired in this way is provided to the target user. It can be suggested to any user, including the target user, to improve their immune status.

Thereby, there is provided a computer program, an information processing apparatus, and a computer program, which are used to estimate the immune status of a target user based on a history of at least one action performed by the target user, and have at least partially improved performance. method can be provided.

(3) Concerning hydration status The estimation model for predicting hydration status is used to predict the severity of symptoms in people whose bodies have reduced immunity. In order to generate such an estimation model, the sample behavior data that is an explanatory variable may be at least one of a plurality of data exemplified below.
・Data that identifies the products (or the category to which the products belong) that the sample user purchased during a certain period of time (for example, one year, etc.) ・The data that identifies the services (or the category to which the services belong) that the sample user used during a certain period of time (for example, one year, etc.) data that identifies categories)

Furthermore, the data identifying the symptom, which is the target variable, may be at least one of a plurality of data (which may be a measured value) exemplified below.
・Data that identifies the sample user's serum Na value ・Data that identifies the sample user's urine osmolarity ・Data that identifies the sample user's urine specific gravity ・Data that identifies the sample user's urine color ・Sample user's BUN/creatinine ratio・Data that identifies the sample user's dehydration evaluation scale ・Data that identifies the sample user's subjective symptoms associated with dehydration

These data may be generated from clinical test values obtained from a medical checkup or medical checkup received by the sample user, or from the contents of the sample user's responses to a health questionnaire. Note that data for identifying the sample user's subjective symptoms associated with dehydration may be generated from the sample user's responses to the questionnaire. The degree of such subjective symptoms can be ascertained from serum Na value, urine osmolality, urine specific gravity, urine color, BUN/creatinine ratio, amount of water consumed, and/or amount of physical activity.

The information processing device can generate an estimation model that predicts the state of hydration by performing machine learning using such explanatory variables and objective variables. The operations performed by the communication system 2 in connection with the generation and utilization method of such an estimation model are described in the above "6" and "7" above in relation to the aspect of predicting the state of menopausal symptoms. The operation may be substantially the same as described. In the explanations given in Sections “6” and “7” above (with particular reference to FIGS. 12 and 14), the explanatory variables and objective variables used in the estimation model for predicting the state of menopausal symptoms are as follows: By replacing them with the explanatory variables and objective variables described in Section 8(3) of this article, the communication system 2 can perform the generation and usage of an estimation model for predicting the state of hydration. The operation will be understood by those skilled in the art.

As explained above, according to the technology disclosed in the present application, by inputting target behavior data that identifies the history of at least one behavior performed by a target user into an estimation model, Target status data may be obtained that identifies the user's current (predicted) hydration status. Furthermore, from among the plurality of target products and/or the plurality of target services prepared in advance, at least one target product and/or at least one target service suitable for the target status data acquired in this way is provided to the target user. Suggestions can be made to any user, including the target user, to improve their hydration status.

Thereby, a computer program, an information processing apparatus, used for estimating the hydration status of a target user based on a history of at least one action performed by the target user, and having at least partially improved performance. and a method can be provided.

(4) Regarding nutritional status , an estimation model for predicting nutritional status is used to predict the severity of symptoms of people who are unable to receive nutritional support. In order to generate such an estimation model, the sample behavior data that is an explanatory variable may be at least one of a plurality of data exemplified below.
・Data that identifies the products (or the category to which the products belong) that the sample user purchased during a certain period of time (for example, one year, etc.) ・The data that identifies the services (or the category to which the services belong) that the sample user used during a certain period of time (for example, one year, etc.) data that identifies categories)

Furthermore, the data identifying the symptom, which is the target variable, may be at least one of a plurality of data (which may be a measured value) exemplified below.
・Data that identifies the presence or absence of subjective symptoms associated with malnutrition in sample users ・Data that identifies sample users' food intake diversity scores (DVS) ・Data that identifies sample users' Simplified Nutritional Appetite Questionnaire (SNAQ) Identification data - Data that identifies the sample user's BMI (Body Mass Index)

Note that the data for identifying such symptoms may be the sample user's subjective symptoms obtained from a questionnaire or the like, or may be obtained from the sample user's reception history at a medical institution. Data identifying such symptoms may identify any of multiple (preferably three or more) categories.

The information processing device can generate an estimation model that predicts nutritional status by performing machine learning using such explanatory variables and objective variables. The operations performed by the communication system 2 in connection with the generation and utilization method of such an estimation model are described in the above "6" and "7" above in relation to the aspect of predicting the state of menopausal symptoms. The operation may be substantially the same as described. In the explanations given in Sections “6” and “7” above (with particular reference to FIGS. 12 and 14), the explanatory variables and objective variables used in the estimation model for predicting the state of menopausal symptoms are as follows: These are executed by the communication system 2 in connection with the generation and utilization method of an estimation model for predicting nutritional status by replacing them with the explanatory variables and objective variables described in Section 8 (4) of this article. Operation will be understood by those skilled in the art.

As explained above, according to the technology disclosed in the present application, by inputting target behavior data that identifies the history of at least one behavior performed by a target user into an estimation model, Target status data may be obtained that identifies the user's current (predicted) nutritional status. Furthermore, from among the plurality of target products and/or the plurality of target services prepared in advance, at least one target product and/or at least one target service suitable for the target status data acquired in this way is provided to the target user. Suggestions can be made to any user, including the target user, to improve their nutritional status.

Thereby, there is provided a computer program, an information processing apparatus, and an information processing apparatus, which are used to estimate the nutritional status of a target user based on a history of at least one action performed by the target user, and have at least partially improved performance. method can be provided.

9. Various Aspects A computer program according to a first aspect is ``executed by at least one processor to obtain target behavior data that identifies a history of behaviors performed by a target user, and to perform supervised learning. Target sleep state data that identifies the sleep state of the target user by inputting the target behavior data into an estimation model generated by the method, and target property data that identifies the sleep quality of the target user; At least one of target rhythm data that identifies the sleep rhythm of the target user, target time data that identifies the sleep time of the target user, and target category data that identifies the sleep category to which the target user belongs. the at least one processor may be configured to cause the at least one processor to output target sleep state data including target sleep state data from the estimation model.
The computer program according to the second aspect is characterized in that in the first aspect, "each set of teacher data includes sample behavior data that identifies the history of actions performed by one sample user corresponding to the set; obtaining the estimated model generated by inputting a plurality of sets of teacher data into a learning model, including sample sleep state data that identifies the sleep state of one sample user corresponding to the set; and causing the at least one processor to function in such a manner.
In the computer program according to the third aspect, in the first aspect, "each set of teacher data includes sample behavior data that identifies the history of actions performed by one sample user corresponding to the set; and sample sleep state data that identifies the sleep state of one sample user corresponding to the set. The at least one processor may be connected via a line.
The computer program according to the fourth aspect is characterized in that in the first aspect, "each set of teacher data includes sample behavior data that identifies the history of actions performed by one sample user corresponding to the set; sample sleep state data that identifies the sleep state of one sample user corresponding to the set; and acquiring a plurality of sets of teacher data, and inputting the plurality of sets of teacher data to a learning model for learning. ``operating the at least one processor to generate the estimated model.''
The computer program according to the fifth aspect is characterized in that in the fourth aspect, "the acquired plurality of sets of teacher data include a first plurality of sets of teacher data; Each set of teacher data includes, as the sample sleep state data, at least sample property data identifying the sleep property of one sample user corresponding to the set, and the sleep state of one sample user corresponding to the set. sample rhythm data for identifying the rhythm of the data, and for each set of teacher data included in the acquired first plurality of sets of teacher data, the sleep characteristics of one sample user corresponding to the set are determined. A sample sleep that identifies a sleep category to which one sample user corresponding to the group belongs based on sample property data that identifies and sample rhythm data that identifies the sleep rhythm of one sample user that corresponds to the group. Preprocessing is performed to generate category data, and each set of teacher data is composed of the sample behavior data of one sample user corresponding to the set and the sample behavior data of one sample user corresponding to the set. The first plurality of sets of preprocessed teacher data including sleep category data are input to the learning model to generate the estimation model, and from the estimation model, as the target sleep state data, The at least one processor may be operated to output target sleep category data identifying a sleep category to which the target user belongs.
A computer program according to a sixth aspect is characterized in that in any of the second to fourth aspects, ``the plurality of sets of teacher data include a second plurality of sets of teacher data; Each set of teacher data included in the teacher data includes, as the sample sleep state data, at least sample property data identifying the sleeping property of one sample user corresponding to the set and one user corresponding to the set. and sample sleep category data that identifies the sleep category to which one sample user corresponding to the group belongs, determined based on the sample rhythm data that identifies the sleep rhythm of the sample user, and from the estimation model. , the at least one processor may be operated to output, as the target sleep state data, target sleep category data that identifies a sleep category to which the target user belongs.
In the computer program according to the seventh aspect, in the fifth aspect or the sixth aspect, "sample property data identifying the sleep property of one sample user corresponding to the group is The score is calculated based on the sample user's answers to a scale, Pittsburgh Sleep Questionnaire, 3D Sleep Scale, or Insomnia Severity Questionnaire, and indicates the sleep of one sample user corresponding to the set. The sample rhythm data that identifies the rhythm of the sample user indicates a social jet lag, which is the absolute value of the difference between the median time of the weekday sleep time zone and the median time of the holiday sleep time zone, of the sample user, and corresponds to the group. The sleep category to which one sample user belongs includes at least one threshold value arranged in correspondence with the score on one of the horizontal axis and the vertical axis, and the other of the horizontal axis and the vertical axis. and at least one threshold value arranged in correspondence with the social jet lag on the axis.
A computer program according to an eighth aspect is characterized in that in any of the second to fourth aspects, "the plurality of sets of teacher data include a third plurality of sets of teacher data; Each set of teacher data included in the teacher data includes, as the sample sleep state data, at least sample property data that identifies the sleep property of one sample user corresponding to the set, and one sample user corresponding to the set. sample rhythm data that identifies the sleep rhythm of a human sample user; and target property data that identifies the sleep quality of the target user as the target sleep state data from the estimation model; The at least one processor may be operated to output target rhythm data for identifying sleep rhythms.
A computer program according to a ninth aspect is characterized in that in any of the second to fourth aspects, ``the plurality of sets of teacher data include a fourth plurality of sets of teacher data; Each set of teacher data included in the teacher data includes, as the sample sleep state data, at least sample property data that identifies the sleep property of one sample user corresponding to the set, and from the estimation model, The at least one processor may be operated to output target property data that identifies the sleep property of the target user as the target sleep state data.
A computer program according to a tenth aspect is a computer program according to any one of the second to fourth aspects, wherein "the plurality of sets of teacher data includes a fifth plurality of sets of teacher data, and the fifth set of teacher data includes a fifth set of teacher data. Each set of teacher data included in the teacher data includes, as the sample sleep state data, at least sample rhythm data that identifies the sleep rhythm of one sample user corresponding to the group, and from the estimation model, The at least one processor may be operated to output target rhythm data that identifies the sleep rhythm of the target user as the target sleep state data.
A computer program according to an eleventh aspect is characterized in that in any of the second to fourth aspects, "the plurality of sets of teacher data include a sixth plurality of sets of teacher data, and the sixth plurality of sets of teacher data include a sixth plurality of sets of teacher data, Each set of teacher data included in the teacher data includes, as the sample sleep state data, at least sample time data that identifies the sleep time of one sample user corresponding to the set, and from the estimation model, the The at least one processor may be operated to output target time data identifying the sleep time of the target user as the target sleep state data.
A computer program according to a twelfth aspect includes, in any one of the first to eleventh aspects, "At least one object corresponding to the target sleep state data among a plurality of products and/or a plurality of services." said at least one processor to determine a product and/or at least one target service.
The computer program according to the thirteenth aspect provides, in the twelfth aspect, “By inputting the target sleep state data into another estimation model generated by executing supervised learning, , causing the at least one processor to output proposal data identifying the at least one target product and/or the at least one target service.
A computer program according to a fourteenth aspect includes, in the twelfth aspect, "inputting the target sleep state data into a search table that associates the target sleep state data with the plurality of products and/or the plurality of services." The at least one processor may be operated to obtain proposal data identifying the at least one target product and/or the at least one target service from the lookup table.
In the computer program according to the fifteenth aspect, in any one of the first to fourteenth aspects, "the at least one processor is a central processing unit (CPU), a microprocessor, and/or may include a graphics processing unit (GPU).
The information processing device according to the sixteenth aspect is provided with the following: “The information processing device includes at least one processor, and the at least one processor acquires target behavior data that identifies a history of actions performed by a target user, and performs supervised learning. Target sleep state data that identifies the sleep state of the target user by inputting the target behavior data into an estimation model generated by executing the target sleep state data, the target property that identifies the sleep quality of the target user. data, target rhythm data that identifies the sleep rhythm of the target user, target time data that identifies the sleep time of the target user, and target category data that identifies the sleep category to which the target user belongs. outputting target sleep state data including one from the estimation model;
Configured like this:
The information processing device according to the seventeenth aspect may be "a terminal device or a server device" in the sixteenth aspect.
In the information processing device according to the eighteenth aspect, in the seventeenth aspect, “the at least one processor is a central processing unit (CPU), a microprocessor, and/or a graphics processing unit (GPU). It can be included.
A method according to a nineteenth aspect is a method performed by at least one processor for executing computer-readable instructions, the at least one processor, by executing the instructions, The sleep state of the target user is determined by a step of acquiring target behavior data that identifies a history of executed actions, and inputting the target behavior data into an estimation model generated by performing supervised learning. Target sleep state data to be identified, including target property data that identifies the sleep quality of the target user, target rhythm data that identifies the sleep rhythm of the target user, and target time data that identifies the sleep time of the target user. , and target category data identifying a sleep category to which the target user belongs, outputting target sleep state data from the estimation model.
In the method according to the twentieth aspect, in the nineteenth aspect, “the at least one processor includes a central processing unit (CPU), a microprocessor, and/or a graphics processing unit (GPU).” I can see it.
A method according to a twenty-first aspect is "a method performed by at least one processor for executing computer-readable instructions, the at least one processor, by executing the instructions, The teacher data includes sample behavior data that identifies the history of actions performed by one sample user corresponding to the group, and sample sleep state data that identifies the sleep state of one sample user corresponding to the group. and target behavior data that identifies a history of actions performed by the target user by inputting the plurality of sets of training data into a learning model and causing it to learn. Target sleep state data that identifies the sleep state of the target user, target property data that identifies the sleep quality of the target user, and target rhythm data that identifies the sleep rhythm of the target user. , target time data that identifies the sleep time of the target user, and target category data that identifies the sleep category to which the target user belongs. and a generation step of generating an estimated model based on the estimated model.
The method according to the twenty-second aspect is the method according to the twenty-first aspect, in which the acquisition step includes the step of acquiring a first plurality of sets of teacher data included in the plurality of sets of teacher data, and Each set of teacher data included in the set of teacher data includes, as the sample sleep state data, at least sample property data that identifies the sleep property of one sample user corresponding to the set, and sample property data corresponding to the set. sample rhythm data for identifying the sleep rhythm of one sample user, and the generation step further includes generating data for each set of teacher data included in the first plurality of sets of teacher data obtained. corresponding to the set based on sample property data that identifies the sleeping property of one sample user corresponding to the set and sample rhythm data that identifies the sleep rhythm of one sample user corresponding to the set. performing a preprocessing step of generating sample sleep category data that identifies the sleep category to which one sample user belongs; data and the sample sleep category data of one sample user corresponding to the set, the first plurality of sets of teacher data after the preprocessing are inputted into the learning model to model the estimation model. and "generating."
The method according to the twenty-third aspect is the method according to the twenty-first aspect, in which “the acquiring step includes a step of acquiring a second plurality of sets of teacher data included in the plurality of sets of teacher data, and the second plurality of sets of teacher data are Each set of teacher data included in the set of teacher data includes, as the sample sleep state data, at least sample property data identifying the sleeping property of one sample user corresponding to the set and one sample user corresponding to the set. and sample rhythm data that identifies a sleep rhythm of a human sample user, and sample sleep category data that identifies a sleep category to which one sample user corresponding to the group belongs, and the second inputting a plurality of sets of training data into the learning model to generate the estimation model.
In the method according to the twenty-fourth aspect, in the twenty-third aspect, "sample quality data identifying the sleep quality of one sample user corresponding to the group includes the Athens Insomnia Scale, the Pittsburgh Sleep Questionnaire, a sample rhythm that indicates a score calculated based on the response content of the sample user to a three-dimensional sleep scale or an insomnia severity questionnaire, and identifies the sleep rhythm of one sample user corresponding to the group; The data indicates the social jet lag, which is the absolute value of the difference between the median time of the weekday sleep period and the median time of the holiday sleep period of the sample user, and one sample user corresponding to the group The sleep category to which it belongs includes at least one threshold value arranged in correspondence with the score on one of the horizontal axis and the vertical axis, and the social jet lag on the other axis of the horizontal axis and the vertical axis. and at least one threshold value arranged in association with the category.
A method according to a twenty-fifth aspect includes, in any one of the twenty-first to twenty-fourth aspects, "by inputting target behavior data that identifies the history of actions performed by the target user into the estimation model, The method may further include an estimating step of outputting target sleep state data for identifying the sleep state of the target user from the estimation model.
In the method according to the twenty-sixth aspect, in any one of the twenty-first to twenty-fifth aspects, "the at least one processor is a central processing unit (CPU), a microprocessor, and/or a graphic may include a processing unit (GPU).
The computer program according to the twenty-seventh aspect is “generated by being executed by at least one processor to obtain target behavior data that identifies the history of actions performed by the target user, and performing supervised learning.” Target menopausal state data that identifies the state of the menopausal symptoms of the target user by inputting the target behavioral data into the estimated model, which is target comprehensive that identifies the comprehensive state of the menopausal symptoms of the target user. A target menopausal state including at least one of state data, target first state data identifying the state of the first symptom of the target user, and second target state data identifying the state of the second symptom of the target user. the at least one processor to cause data to be output from the estimation model.
The computer program according to the twenty-eighth aspect is characterized in that in the twenty-seventh aspect, “each set of teacher data includes sample behavior data identifying the history of actions performed by one sample user corresponding to the set; and sample menopausal state data that identifies the state of menopausal symptoms of one sample user corresponding to the set, and obtains the estimated model generated by inputting a plurality of sets of teacher data into a learning model. , causing the at least one processor to function as described above.
The computer program according to the twenty-ninth aspect is characterized in that in the twenty-seventh aspect, “each set of teacher data includes sample behavior data identifying the history of actions performed by one sample user corresponding to the set; For the estimation model generated by inputting a plurality of sets of teacher data into a learning model, including sample menopausal state data that identifies the state of menopausal symptoms of one sample user corresponding to the set, said at least one processor may be connected via a communication line.
The computer program according to the 30th aspect is characterized in that in the 27th aspect, "each set of teacher data includes sample behavior data identifying the history of actions performed by one sample user corresponding to the set; Obtain multiple sets of teacher data, including sample menopausal status data that identifies the state of menopausal symptoms of one sample user corresponding to the set, and input the multiple sets of teacher data into a learning model for learning. the at least one processor may be operated to generate the estimated model.
The computer program according to the 31st aspect is characterized in that in the 30th aspect, ``the acquired plurality of sets of teacher data include a first plurality of sets of teacher data; Each set of training data includes, as the sample menopausal state data, at least sample comprehensive state data that identifies the comprehensive state of menopausal symptoms of one sample user corresponding to the set, and from the estimation model, Outputting target comprehensive state data that identifies the comprehensive state of menopausal symptoms of the target user as the target menopausal state data;
and causing the at least one processor to function in such a manner.
The computer program according to the 32nd aspect is characterized in that in the 31st aspect, "Each set of teacher data included in the first plurality of sets of teacher data corresponds to at least the set as the sample menopausal state data. Contains sample comprehensive state data for identifying a score corresponding to the comprehensive state of menopausal symptoms of one sample user, and corresponds to each set of teacher data included in the first plurality of sets of teacher data. The sample overall status data that identifies the score corresponding to the overall state of menopausal symptoms of one sample user corresponding to the group, and the category corresponding to the overall state of menopausal symptoms of one sample user corresponding to the set. Executing preprocessing of converting into sample comprehensive state data to be identified, inputting the first plurality of sets of teacher data after the preprocessing to the learning model to generate the estimation model, and generating the estimation model. The at least one processor may be operated to output, as the target menopausal state data, target comprehensive state data that identifies a category corresponding to the comprehensive state of menopausal symptoms of the target user. .
In the computer program according to the thirty-third aspect, in the thirty-second aspect, "The score corresponding to the comprehensive state of menopausal symptoms of one sample user corresponding to the group is the Simplified Menopausal Index (SMI). , Menopause Rating Scale, Kupperman index, Green Climatic Scale, PSST (Premenstrual Symptoms Screen) ing Tool), WHQ (Women's Health Questionnaire), VAS (Visual Analogue Scale) ), HFRDI (Hot Flash Related Daily Interference Scale), HFCS (Hot Flash Composite Score), MENQOL (Menopause-Specific Quality Of Calculated based on the sample user's responses to the Japanese Women's Menopausal Symptom Evaluation Sheet. It can be determined by the score.
A computer program according to a thirty-fourth aspect is the computer program according to the thirty-fourth aspect, in which the acquired plurality of sets of teacher data includes a second plurality of sets of teacher data, and the second plurality of sets of teacher data include Each set of training data includes, as the sample menopausal state data, at least sample first state data that identifies the state of the first symptom of one sample user corresponding to the set, and from the estimation model, the The at least one processor may be operated to output target first state data that identifies the state of the first symptom of the target user as the target menopausal state data.
The computer program according to the thirty-fifth aspect is characterized in that in the thirty-fourth aspect, "Each set of teacher data included in the second plurality of sets of teacher data corresponds to at least the set as the sample menopausal status data. Contains sample first state data that identifies a score corresponding to the state of the first symptom of one sample user, and corresponds to each set of teacher data included in the second plurality of sets of teacher data. The sample first state data identifying the score corresponding to the first symptom state of one sample user is added to the sample first state data identifying the category corresponding to the first symptom state of one sample user corresponding to the set. Perform preprocessing to convert into one-state data, input the preprocessed second plurality of sets of teacher data to the learning model to generate the estimated model, and from the estimated model, The at least one processor may be operated to output target first state data identifying a category corresponding to a first symptom state of the target user as target menopausal state data.
A computer program according to a thirty-sixth aspect is the computer program according to the thirty-sixth aspect, in which “the acquired plurality of sets of teacher data includes a third plurality of sets of teacher data; Each set of training data includes, as the sample menopausal state data, at least sample second state data that identifies the state of the second symptom of one sample user corresponding to the set, and from the estimation model, the The at least one processor may be operated to output target first state data that identifies the state of the second symptom of the target user as the target menopausal state data.
The computer program according to the thirty-seventh aspect is the computer program according to the thirty-sixth aspect, in which “each set of teacher data included in the third plurality of sets of teacher data corresponds to at least the set as the sample menopausal state data. including sample second state data for identifying a score corresponding to the state of the second symptom of one sample user, and corresponding to each set of teacher data included in the third plurality of sets of teacher data; The sample second state data identifying the score corresponding to the second symptom state of one sample user is added to the sample second state data identifying the category corresponding to the second symptom state of one sample user corresponding to the set. Perform preprocessing to convert into two-state data, input the preprocessed third plurality of sets of teaching data to the learning model to generate the estimated model, and from the estimated model, The at least one processor may be operated to output target second state data that identifies a category corresponding to the second symptom state of the target user as the target menopausal state data.
A computer program according to a thirty-eighth aspect includes, in any one of the twenty-seventh to thirty-seventh aspects, "At least one object corresponding to the target menopausal state data among a plurality of products and/or a plurality of services." said at least one processor to determine a product and/or at least one target service.
The computer program according to the thirty-ninth aspect is, in the thirty-eighth aspect, “by inputting the target menopausal state data into another estimation model generated by executing supervised learning, , causing the at least one processor to output proposal data identifying the at least one target product and/or the at least one target service.
The computer program according to the fortieth aspect includes, in the thirty-eighth aspect, "inputting the target menopausal state data into a search table that associates the target menopausal state data with the plurality of products and/or the plurality of services." The at least one processor may be operated to obtain proposal data identifying the at least one target product and/or the at least one target service from the lookup table.
In the computer program according to the forty-first aspect, in any of the twenty-seventh to fortieth aspects, "the at least one processor is a central processing unit (CPU), a microprocessor, and/or may include a graphics processing unit (GPU).
The information processing device according to the forty-second aspect is provided with the following: “The information processing device includes at least one processor, and the at least one processor acquires target behavior data that identifies a history of actions performed by a target user, and performs supervised learning. Target menopausal state data that identifies the state of menopausal symptoms of the target user by inputting the target behavior data into an estimation model generated by executing the target user, the comprehensive state of the menopausal symptoms of the target user. at least the following: target comprehensive state data for identifying the target user's first symptom state; and target second state data for identifying the target user's second symptom state. The estimation model may be configured to output target menopausal state data including one item from the estimation model.
The information processing device according to the forty-third aspect may be "a terminal device or a server device" in the forty-second aspect.
In the information processing device according to the forty-fourth aspect, in the forty-second aspect or the forty-third aspect, “the at least one processor is a central processing unit (CPU), a microprocessor, and/or a graphics processor. may include a processing unit (GPU).
A method according to a forty-fifth aspect is “a method performed by at least one processor that executes computer-readable instructions, the at least one processor, by executing the instructions, The state of menopausal symptoms of the target user is determined by a step of acquiring target behavior data that identifies a history of performed actions, and inputting the target behavior data into an estimation model generated by performing supervised learning. target comprehensive state data that identifies the comprehensive state of menopausal symptoms of the target user, target first state data that identifies the state of the first symptom of the target user, and outputting target menopausal state data including at least one of target second state data identifying the state of the second symptom of the target user from the estimation model.
In the method according to the forty-sixth aspect, in the forty-fifth aspect, "the at least one processor includes a central processing unit (CPU), a microprocessor, and/or a graphics processing unit (GPU)." I can see it.
A method according to a forty-seventh aspect is "a method performed by at least one processor for executing computer-readable instructions, the at least one processor, by executing the instructions, Sample behavior data in which the teacher data identifies a history of actions performed by one sample user corresponding to the group, and a sample menopausal state that identifies the state of menopausal symptoms of one sample user corresponding to the group. a target action that identifies a history of actions performed by the target user by inputting the plurality of sets of training data into a learning model and making it learn; Target menopausal state data for identifying the state of menopausal symptoms of the target user by inputting data, the target comprehensive state data identifying the comprehensive state of menopausal symptoms of the target user; Output target menopausal state data including at least one of target first state data that identifies a symptom state and target second state data that identifies a second symptom state of the target user. a generation step of generating a constructed estimation model.
The method according to the forty-eighth aspect is the method according to the forty-seventh aspect, in which the acquisition step includes the step of acquiring a first plurality of sets of teacher data included in the plurality of sets of teacher data, and Each set of teacher data included in the set of teacher data includes, as the sample menopausal state data, at least a sample composite that identifies a score corresponding to the overall state of menopausal symptoms of one sample user corresponding to the set. The generation step further includes, for each set of teacher data included in the acquired first plurality of sets of teacher data, a comprehensive list of menopausal symptoms of one sample user corresponding to the set. sample general state data that identifies a score corresponding to a condition, into sample comprehensive state data that identifies a category corresponding to a comprehensive state of menopausal symptoms of one sample user corresponding to the set. a step of performing processing, and each set of teacher data includes the sample behavior data of one sample user corresponding to the set and the sample comprehensive state data of one sample user corresponding to the set. inputting the first plurality of sets of teacher data after the preprocessing into the learning model to generate the estimation model.
The method according to the forty-ninth aspect is the method according to the forty-seventh aspect, in which “the acquisition step includes the step of acquiring a first plurality of sets of teacher data included in the plurality of sets of teacher data, and the second plurality of sets of teacher data are Each set of teacher data included in the set of teacher data includes at least a sample first state that identifies, as the sample menopausal state data, a score corresponding to the state of the first symptom of one sample user corresponding to the set. further, the generating step generates a first symptom state of one sample user corresponding to the set for each set of teacher data included in the acquired second plurality of sets of teacher data. Perform preprocessing of converting sample first state data identifying the corresponding score into sample first state data identifying the category corresponding to the first symptom state of one sample user corresponding to the set. each set of teacher data includes the sample behavior data of one sample user corresponding to the set, and the sample first state data of one sample user corresponding to the set; inputting the second plurality of sets of teacher data after the preprocessing into the learning model to generate the estimation model.
The method according to the fiftieth aspect is the method according to the forty-seventh aspect, in which “the acquisition step includes the step of acquiring a third plurality of sets of teacher data included in the plurality of sets of teacher data, and the third plurality of sets of teacher data are Each set of teacher data included in the set of teacher data includes a sample second state that identifies, as the sample menopausal state data, at least a score corresponding to the state of the second symptom of one sample user corresponding to the set. further, the generating step generates a state of a second symptom of one sample user corresponding to the set for each set of teacher data included in the acquired third plurality of sets of teacher data. Preprocessing is performed to convert the sample second state data identifying the corresponding score into sample second state data identifying the category corresponding to the second symptom state of one sample user corresponding to the set. each set of teacher data includes the sample behavior data of one sample user corresponding to the set, and the sample second state data of one sample user corresponding to the set; inputting the third plurality of sets of teacher data after the preprocessing into the learning model to generate the estimation model.
In the method according to the 51st aspect, in the 47th aspect to the 50th aspect, "the at least one processor is a central processing unit (CPU), a microprocessor, and/or a graphics processing unit. (GPU).

II. Chapter 2 It is possible to use a combination of the technology described in Chapter 1 and the technology described below in Chapter 2.

For example, the estimation model for estimating the sleep state described in Chapter 1 can be used as the sleep state prediction model 110b in the technology described below. Similarly, the estimation model for estimating the state of menopausal symptoms described in Chapter 1 can be used as the female health prediction model 100a in the technology described below. Furthermore, similarly, the estimated model that predicts the exercise status, the estimated model that predicts the immune status, the estimated model that predicts the hydration status, and the estimated model that predicts the nutritional status, which were explained in Chapter 1. , in the techniques described below, can be used as a lifestyle-related disease prediction model 110c, an immune status prediction model 110e, a moisture status prediction model 110d, and a nutritional status prediction model 110f, respectively.

Without limitation, at least one feature included in the technology explained in Chapter 1 (or at least one feature included in the technology explained in Chapter 2) can be replaced by the technology explained in Chapter 2 (or at least one feature included in the technology explained in Chapter 1). It is also possible to use it in the technology described in .

1 Overview FIGS. 15 to 28 are diagrams for explaining the embodiment. In the following, we will broadly explain a health prediction system that predicts a customer's health condition (also referred to as a health symptom), and a learning system that generates a predictive model for predicting a customer's health condition, which is used in the health prediction system. The overview, functional configuration, processing flow, etc. will be explained in sequence.

2 Health prediction system
2.1 Outline of Processing of Health Prediction System First, a rough outline of processing of the health prediction system 1Z according to the embodiment will be explained with reference to FIG. FIG. 15 schematically shows the flow of data for predicting a customer's health using the health prediction system 1Z according to the embodiment.

The health prediction system 1Z is for estimating an individual's health condition from the individual's purchasing behavior. By using this health prediction system 1Z, the health condition of an individual customer can be estimated based on purchase data, which is the customer's purchasing behavior, without data related to the customer's health such as medical examination data. The estimated health status can be used, for example, to improve the customer's health.

Here, the integrated health prediction program 100Z that predicts the health condition of an individual based on the individual's purchasing behavior includes a plurality of types of learned prediction models that predict the health condition from the individual's purchasing behavior. In the example of FIG. 15, the integrated health prediction program 100Z is configured by integrating six prediction models, but it is not limited to six, and may be five or less or seven or more. It's okay.

The integrated health prediction program 100Z according to the present embodiment includes a female health prediction model 110a that predicts women's health, a sleep state prediction model 110b that predicts sleep states, a lifestyle disease prediction model 110c that predicts lifestyle diseases, and water. There are six prediction models: a moisture status prediction model 110d that predicts the state, an immune status prediction model 110e that predicts the immune status, and a nutritional status prediction model 110f that predicts the nutritional status (hereinafter collectively referred to as the "prediction model 110Z"). including. Each prediction model included in the integrated health prediction program 100Z calculates a woman's health status (SMI related to menopausal symptoms, etc.) / sleep status / lifestyle-related diseases / hydration status (hydration status) from purchase data 150Z that shows the customer's purchase history. A value indicating /immune status/nutritional status (a value indicating the possibility and/or degree of suffering from a specific disease) is calculated as the customer's predicted health value 160Z. In the example of FIG. 15, values indicating six health conditions are shown in percentages. Depending on the type of disease, only men, only women, or both men and women may be targeted. These predictive models 110Z may be described as functions. It is built using machine learning using training data that includes product purchase history and various health-related conditions. The method for constructing the prediction model will be described later.

More specifically, the integrated health prediction program 100Z receives customer purchasing data 150Z. The purchase data 150Z is data related to the purchase history of a customer whose health is to be predicted, and corresponds to the explanatory variables of the teacher data at the time of constructing each prediction model 110Z included in the integrated health prediction program 100Z. The integrated health prediction program 100Z inputs purchase data 150Z into each prediction model 110Z in parallel, and calculates a customer's health prediction value 160Z. In this embodiment, only the purchase data 150Z is input to the prediction model 110Z, but customer attribute information such as the customer's gender and age may also be input, for example.

Further, the integrated health prediction program 100Z may make various proposals regarding health improvement etc. to the customer according to the calculated health prediction value 160Z. More specifically, for example, it is possible to point out to the customer categories with high health prediction values and suggest measures that can be taken while the customer is unaware of (latent) symptoms or symptoms that have become apparent. Conceivable. The proposed method will be described later with reference to FIG. 18 and the like.

2.2 Configuration of Integrated Health Prediction Program The functional configuration of the integrated health prediction program 100Z will be described below with reference to FIG. 16. FIG. 16 is a diagram showing the functional configuration of the integrated health prediction program 100Z. The integrated health prediction program 100Z includes an input section 120Z, a product category determination section 125Z, a prediction model 110Z, an output section 130Z, and a suggestion section 135Z. Note that the integrated health prediction program 100Z may be implemented as a single program, or may be implemented as a plurality of programs that can operate in cooperation by inputting and outputting data.

The input unit 120Z receives (obtains) input of purchase data 150Z. The input unit 120Z may read the purchase data 150Z from a storage medium such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive) built in the information processing device (computer) on which the integrated health prediction program 100Z is executed. Alternatively, the purchase data 150Z may be input from an external information processing device connected via a LAN (Local Area Network), the Internet, or the like.

Here, with reference to FIG. 17, a specific example of the purchase data 150Z that the input unit 120Z receives will be described. FIG. 17 is a diagram showing a specific example of purchase data 150Z. Here, the purchase data 150Z may be panel data provided by a research company that accumulates customer purchase history. The panel data is data obtained by scanning the product by a customer who is the purchaser when the product is purchased. When a customer scans a product using his or her smartphone or a loaned barcode reader, the product is associated with preset customer attributes (office worker/student/housewife, etc.), purchaser information, etc. is accumulated. This allows us to understand what kind of person purchased what, how much, and at what kind of store.

The example of purchase data 150Z in FIG. 17 shows the purchase history of food and/or daily necessities purchased by a customer over a certain period of time (for example, one year). Each product purchased by a customer is classified into a product category in the JICFS (JAN Item Code File Service) classification. In other words, each product category includes at least one product. Purchasing data 150Z includes approximately 500 items (for example, soft drinks, soft drinks, Each item 150a (mouthwash, etc.) is entered in association with performance data 150b, which is numerical data such as purchase amount, purchase quantity, and number of purchases during the period.

In this embodiment, the purchase data 150Z includes the purchase history of food and/or daily necessities, but it is not limited to this as long as it is presumed to be related to the customer's health condition. . For example, the purchase data 150Z may include purchase history of other products such as medicines and various services such as massage and acupuncture. Further, the purchase data 150Z may be specified using payment information managed by a retail store, or may be specified using, for example, credit card payment information, payment information to a financial institution, etc.

Upon receiving the purchase data 150Z from the input unit 120Z, the product category determining unit 125Z determines the classification code (JICFS classification) and product category to be input to each prediction model 110Z from the purchase data 150Z input to each prediction model 110Z, Performance data 150b, which is numerical data such as purchase amount, purchase quantity, and number of purchases regarding the obtained classification code and product category (only a part of these may be used), is input in parallel to each prediction model 110Z. do.

Note that the classification code (product category) of the purchase data 150Z input to the prediction model 110Z can be determined for each prediction model 110Z. A method for determining the product category to be input to the prediction model 110Z will be described later in 3.4 below regarding creation of the prediction model 110Z.

Note that it is also conceivable that the purchase data 150Z input from the input unit 120Z is inputted as is into the prediction model 110Z without narrowing down the purchase data 150Z by product category. In that case, the product category determining section 125Z is unnecessary.

The prediction model 110Z receives input of performance data 150b, which is numerical data such as purchase amount, purchase quantity, and number of purchases, regarding the classification code and product category selected by the product category determination unit 125Z, and receives input of performance data 150b, which is numerical data such as purchase amount, purchase quantity, and number of purchases, and determines the health status of the woman (menopause). Symptoms), sleep status (disturbed sleep), lifestyle-related diseases (lack of physical activity), hydration status (degree of hydration/hydration status), immune status (degree of immune deficiency), and nutritional status (degree of nutritional deficiency) Calculate the predicted value. In the example of FIG. 15, for the customer with user ID 001, each prediction model 110Z has 90% of menopausal symptoms, 21% of disturbed sleep, 40% of physical inactivity, 76% of hydration, 72% of decreased immunity, The nutritional deficiency level is calculated at 11%.

As mentioned earlier, the predictive model 110Z uses the purchasing data 150Z, which is the purchasing history of each customer, and the training data that correlates the health status of those customers, and performs machine learning using the former as an explanatory variable and the latter as an objective variable. It can be constructed by The method for constructing the prediction model 110Z will be described later in Section 3 below.

The output unit 130Z outputs the values calculated by each prediction model 110Z as a predicted health value 160Z for the customer to the proposal unit 135Z, a built-in storage medium, or another information processing device connected via a network or the like. Output. Alternatively, the output unit 130Z may display the calculated predicted health value 160Z on a display device. In the example of FIG. 15, the output unit 140Z arranges values such as customer ID, menopause, sleep disorder, etc. in order in the row direction, and repeats this for each ID of customer data to predict health for multiple customers. The value 160Z is output. Note that the predicted health value 160Z may include not only the customer's actual health condition but also a health condition that can be expected to occur in the future (potential health condition).

The proposal unit 135Z proposes foods, supplies, services, etc. according to the predicted health prediction value 160Z for each customer. There are various ways to make proposals, but for example, it may be displayed on various terminals used by retail store staff, the proposal may be sent directly to the customer via a messenger service, or the proposal may be posted on a web page to the customer. It is also possible to display the information on

2.3 Specific Examples and Processing of Various Proposals to Customers Various proposals that the integrated health prediction program 100Z makes to customers will be described below with reference to FIGS. 18 to 20.

FIG. 18 is a diagram schematically showing foods/goods/services etc. proposed by the suggestion unit 135Z according to the customer's predicted health value 160Z. For example, for a woman whose menopausal symptoms are predicted to be 90% (customer ID=001), the suggestion unit 135Z suggests products that improve menopausal symptoms that are sold at the drug store 41Z. Further, for a man (customer ID=002) whose degree of nutritional deficiency is predicted to be 93%, the proposal unit 135Z proposes a highly functional nutritional food sold at the drug store 41Z. As another example, the proposal unit 135Z suggests a visit to the medical institution 43Z to a woman (customer ID=003) whose predicted values for menopause and sleep disturbance are 79% and 89%, respectively. Further, the proposal unit 135Z suggests that customers whose degree of lack of physical activity is relatively high should go to the gym 45Z, and for customers whose degree of lack of physical activity is lower than a predetermined value, the insurance company 47Z provides , proposes medical insurance products that address lifestyle-related diseases.

In this way, the proposal unit 135Z of the integrated health prediction program 100Z can propose foods/goods/services suitable for the customer by obtaining prediction results from each of the plurality of prediction models 110Z from the purchasing data 150Z. can. To implement such a proposal, a list of foods, supplies, and/or services associated with each health symptom, which is an output value of a predictive model, may be prepared in advance.

FIG. 19 shows a table 50Z and databases 52Z, 54Z, and 56Z to be referred to for realizing suggestions regarding menopausal symptoms.

Table 50Z includes three classes, class 50a where the output range of the menopause prediction model (female health prediction model 110a in FIG. 16) is less than 40%, class 50b where the output range is 40% or more and less than 80%, and class 50c where the output range is 80% or more. There are two classes. In the class 50a, a drugstore product database 52Z for preventing menopausal symptoms is referenced. In the product database 52Z, products X1 and X2 of product category P1 and product X3 of product category P2 are described. The suggestion unit 135Z can refer to the product database 52Z and suggest some or all of the categories and/or product names to the corresponding customer. When the output value of the prediction model corresponds to the class 50b, the proposal unit 135Z refers to the product database 54Z for improving menopausal symptoms and proposes products Y1 and Y2 of the product category Q1. When the output value of the prediction model corresponds to 50c, the proposal unit 135Z refers to the menopausal outpatient medical institution database 56Z using the current location and/or the pre-registered customer residence information and proposes a medical institution. In other words, this corresponds to introducing the customer to a medical institution that provides preventive medical care and/or treatment according to the customer's health prediction results.

Another example will be described with reference to FIG. 20. FIG. 20 shows a table 60Z for realizing proposals according to the output values of the lifestyle-related disease prediction model 110c.

The table 60Z shows the output values of the lifestyle-related disease prediction model 110c, in other words, a class 60a where the range of the degree of lack of physical activity is less than 20%, a class 60b where the range is 20% or more and less than 80%, and a class where the range is 80% or more. There are three classes of 60c. In class 60a, physical activity is considered to be substantial, and in order to encourage students to continue doing physical activity, nutritional guidance (1) and a monthly membership fee plan for an E1 sports gym are proposed. Nutritional guidance (1) is, for example, a meal suggestion for obtaining nutritional components of less than 10 grams of fat and 30 grams or more of protein per meal. In class 60b, it is considered that there is no shortage of physical activity, and in order to provide nutritional guidance (2) and to regularly engage in physical activity, a pay-as-you-go membership plan at the E2 fitness club is proposed. Nutritional guidance (2) is, for example, a proposal for a meal to provide nutritional components of less than 20 grams of fat and 20 grams or more of protein per meal. In class 60c, physical activity is insufficient and there is a possibility of contracting a lifestyle-related disease in the future, so nutritional guidance (3) and medical insurance products proposed by E3 Life Insurance Company are recommended. Nutritional guidance (3) is, for example, a meal suggestion for ingesting nutritional components of less than 20 grams of fat, 20 grams or more of protein, and less than 20 grams of carbohydrates per meal. Proposals for nutritional guidance, gyms, fitness clubs, and insurance products fall under the category of product or service provision. Regarding the output values of the lifestyle-related disease prediction model 110c, the proposal unit 135Z can make proposals to the customer according to the output values by referring to the table 60Z.

In addition to the examples shown in FIGS. 19 and 20, by providing similar tables and databases as necessary for each of the sleep state, immune state, hydration state, and nutritional state, the proposal unit 135Z can Proposals can be made to the customer according to the predicted health value 160Z, which is the output value of the predictive model 110Z.

Note that the integrated health prediction program 100Z itself does not necessarily need to include the proposal section 135Z. For example, the output unit 130Z of the integrated health prediction program 100Z transmits the predicted health value 160Z to a business other than the one that operates the integrated health prediction program 100Z with the customer's permission, and It is also conceivable that the program of the business implements the program of the proposal section 135Z and makes the above-mentioned proposal to the customer. For example, the integrated health prediction program 100Z may send the health prediction value 160Z, which is the output value of the prediction model 110Z, to an insurance company, so that the insurance company may propose an insurance product to the customer.

2.4 Processing Flow The processing flow of the integrated health prediction program 100Z will be explained below with reference to FIG. FIG. 21 is a flowchart showing the procedure of a customer's health prediction process executed by the integrated health prediction program 100Z.

The input unit 120Z of the integrated health prediction program 100Z acquires purchase data 150Z indicating the purchase history of a customer whose health condition is to be predicted (S701). For example, when the customer purchases a product at a retail store, the input unit 120Z reads the customer's membership card from a POS (Point of Sale) system, thereby constructing a database that links the customer with the purchase history. By extracting the purchase data 150Z from such a database with the consent of the customer, the input unit 120Z can input the purchase data 150Z into the integrated health prediction program 100Z. As mentioned above, such purchase data 150Z includes, for example, JICFS classification and product category.

The product category determination unit 125Z determines the JICFS classification (classification code) and product category to be input to each prediction model 110Z from the purchase data 150Z (S703). Note that, as described above, if the prediction model 110Z does not particularly narrow down the product categories used for prediction and uses the entire purchase data 150Z, the process of S703 is not necessary.

Next, each prediction model 110Z predicts the health condition using the classification code determined by the product category determination unit 125Z and the purchase data 150Z of the product category (S705). The output unit 130Z outputs a predicted health value 160Z indicating the health condition predicted by the predictive model 110Z, and the proposal unit 135Z provides the customer with a proposal based on the predicted health value 160Z (S707). Specifically, it is conceivable that the predicted health value 160Z or the suggestion unit 135Z proposes health improvement to the customer in the form of printing on a receipt, displaying it on a display, notifying a smartphone, or the like.

2.5 Hardware Configuration A specific example of the hardware configuration of the information processing device 800Z on which the integrated health prediction program 100Z is executed will be described with reference to FIG. FIG. 22 is a hardware configuration diagram of an information processing device 800Z on which the integrated health prediction program 100Z is executed. The information processing device 800Z uses the integrated health prediction program 100Z to calculate prediction results regarding the client's menopausal symptoms, sleep state, and the like. The information processing device 800Z may be a generally available computer system, such as a desktop PC (Personal Computer), a notebook PC, a tablet PC, a server computer, or the like. The computer system may be installed at the office of the business operator that operates the health prediction system 1Z, or may be provided as a cloud service. In the latter case, the business office only needs to be equipped with a PC or the like that can communicate with the cloud service.

The information processing device 800Z includes a control section 810Z, an input interface (I/F) section 820Z, a storage device 830Z, and an output I/F section 840Z.

The control unit 810Z may include a CPU (Central Processing Unit, not shown), a ROM (Read Only Memory, not shown), a RAM (Random Access Memory), and the like. The control unit 810Z can execute the integrated health prediction program 100Z stored in the storage device 830Z. Thereby, the information processing device 800Z is capable of performing various processes related to the health prediction described above in addition to the functions of a general computer. Note that the arithmetic circuit included in the control unit 810Z does not need to be a CPU, and may be realized by various processors such as an MPU or a GPU, or may be realized by a plurality of processors instead of one processor.

The input I/F unit 820Z receives purchase data 150Z regarding a customer whose health condition is desired to be predicted. As described above, the input I/F unit 820Z can receive the purchase data 150Z from an information processing device such as an external server connected via a network such as the Internet or a LAN. The input I/F unit 820Z can be realized by, for example, an Ethernet (registered trademark) communication terminal, a USB (registered trademark) terminal, a communication circuit that performs communication in accordance with standards such as IEEE802.11, 4G, or 5G. can.

The storage device 830Z is a storage medium that stores computer programs and data necessary for operating the information processing device 800Z. The storage device 830Z may be, for example, an HDD or an SSD that is a semiconductor storage device. The storage device 830Z may include a temporary storage element constituted by a RAM such as a DRAM or an SRAM, and may function as a work area for the control unit 810Z. The storage device 830Z stores the integrated health prediction program 100Z and a proposal table 831Z, specific examples of which are shown as the table 50Z in FIG. 19 and the table 60Z in FIG. 20.

Note that the prediction model 110Z may be incorporated as a part of the integrated health prediction program 100Z, or may be provided as data separate from the integrated health prediction program 100Z. Further, the prediction model 110Z may be stored in the storage device 830Z as a table or a parameter group.

In addition, the proposal table 831Z shows the prediction results indicating the degree of health condition, for example, H layer/M layer/L layer or 90%/50%/20%, and products/services to be proposed according to each degree. It can be made to correspond.

The output I/F unit 840Z is a communication circuit and/or a communication terminal connected to various output devices provided outside the information processing device 800Z. For example, the output I/F unit 840Z may be a USB terminal that outputs print data indicating the content to be printed on the receipt printer P. The integrated health prediction program 100Z can calculate a prediction result regarding the customer's health condition, and send the result or a proposal to the customer according to the result to the receipt printer P to print it on a receipt.

The output I/F section 840Z may be a video output terminal connected to the display D. The integrated health prediction program 100Z calculates a prediction result regarding the customer's health condition, and causes the display D to display the result or a proposal to the customer according to the result. For example, by checking the contents displayed on the display D by a pharmacist, the pharmacist can suggest one or more products according to the customer's symptoms.

As another form, the output I/F unit 840Z may be a communication terminal or a communication circuit that can be connected to the communication network N or the like and perform data communication. When the output I/F section 840Z is a communication terminal or a communication circuit, the hardware of the input I/F section 820Z and the output I/F section 840Z may be the same. The output I/F unit 840Z can transmit the prediction result regarding the customer's health condition or a proposal to the customer according to the result to the customer's smartphone M, for example, via a mobile phone line.

3 Learning system 9 for learning prediction model 110Z
3.1 Overview Next, a method for generating the prediction model 110Z included in the integrated health prediction program 100Z by machine learning will be described with reference to FIG. 23. Here, the explanation will be given assuming that the model learning program 900Z generates the predictive model 110Z. Note that the model learning program 900Z may be implemented as a single program, or as a plurality of programs that can operate in cooperation by inputting and outputting data.

Note that the integrated health prediction program 100Z shown in FIG. can be generated. In the following, a case will be mainly described in which a menopausal prediction model 110a related to women's health is generated.

3.2 Teacher data 950Z
The model learning program 900Z performs machine learning using teacher data 950Z including various data collected from a large number of subjects, and constructs a predictive model 110Z. The teacher data 950Z includes purchase data 952Z regarding the purchasing behavior of each subject and health data 954Z that is data regarding the health of those subjects.

The purchase data 952Z is related to the same information as the specific example shown in FIG. 17. That is, in this embodiment, for each subject ID, items 150a related to categories in the JICFS classification and performance data 150b, which is numerical data such as purchase amount, purchase quantity, and number of purchases during the period, are included in purchase data 952Z. can be included. Note that since foods and/or drinks are suitable as products that are associated with the health conditions of test subjects and customers, it is preferable that the purchase data 952Z includes the purchase history of the foods and/or drinks.

Health data 954Z describes the health condition of the subject. This health condition may be a condition related to subjective symptoms or a condition recognized as a result of measurement. That is, it is possible to use at least one of health consciousness, mental health, cognitive function, and health checkup results as the health data 954Z. In the health data 954Z in this embodiment, a determination value for each condition segment of the subject is described as the subject's health condition. The condition segment is defined for each combination of health condition type and level (in other words, size). The type and level of health condition in this embodiment refers to the type and level of physical and mental illness that the corresponding individual has. The determination value can be expressed as an index indicating the health condition in 2 classes, 3 classes, or the like. Note that the "type and level of health condition" referred to here is not accompanied by medical rigor, but can also be understood as the type and level of health concerns that an individual has.

Here, as an example, consider a case where the model learning program 900Z generates the menopause prediction model 110a and uses symptoms of the simplified menopause index SMI to determine the level of menopause. In this case, the health data 954Z can include a label indicating the symptoms of the simplified menopause index SMI to which the subject belongs. For example, it is possible to express the simplified menopause index SMI as two classes. Specifically, "H" indicates that menopausal symptoms are relatively large, and "other than H". "Other than H" can be subdivided into "M", which indicates relatively moderate symptoms, and "L", which indicates relatively small symptoms (i.e., symptoms can be expressed as a total of three classes). It is. Hereinafter, the subject layer corresponding to "H" will be referred to as the "H layer."

In addition, in labeling to indicate the degree of symptoms in Health Data 954Z, instead of 2 or 3 classes, it may be further subdivided into 4 or more classes, or instead of these classes, it may be possible to label symptoms that correspond to menopausal symptoms. It is also possible to use numerical values indicating gender.

That is, the teacher data 950Z includes at least purchase data 952Z including information regarding each product category of each subject, information regarding purchase amount, purchase quantity, number of purchases, etc. during a predetermined period, and information on labels indicating health conditions. The health data 954Z that includes the information is linked. Here, the model learning program 900Z can be used to determine, for example, a woman's health condition (menopausal symptoms), sleep condition (disturbed sleep), lifestyle-related diseases (lack of physical activity), hydration condition (degree of water deficiency), and immune condition (low immunity). If six predictive models 110Z for nutritional status (degree) and nutritional status (degree of malnutrition) are to be generated, information corresponding to each of these predictive models 110Z may be included in the health data 954Z. In machine learning in supervised learning, the teacher data 950Z is broadly classified into explanatory variables and objective variables. In this embodiment, the purchase data 952Z corresponds to an explanatory variable, and the health data 954Z corresponds to an objective variable. Note that in this embodiment, only the purchase data 952Z is used as the explanatory variable, but the explanatory variable is not limited to this. For example, it is conceivable to add gender and age, which are attributes of the subject, to explanatory variables.

Below, the model learning program 900Z can be used to improve sleep status, lifestyle-related diseases, immune status, hydration status, nutritional status, dementia, oxygen utilization, vascular health, hair health, and intestinal health by preparing appropriate health data 954Z. Environment, promoting exercise effects, increasing appetite, eye health, skin health, oral health, mental health, depression, stress, headaches, stiff shoulders, ear health, joint health, body temperature homeostasis, fatigue, bone health It is possible to generate predictive models 110Z regarding various symptoms such as male menopausal disorder, high blood pressure, and weather pain. In particular, we will focus on women's health status (menopausal symptoms), sleep status (disturbed sleep), lifestyle-related diseases (lack of physical activity), hydration status (degree of dehydration), immune status (degree of immune deficiency), and nutritional status. The following is an example of an index that can be used as an indicator of (degree of nutritional deficiency).

3.2.1 Women's health status (menopausal symptoms)
As an index indicating the degree of menopausal symptoms, for example, as described above, the simplified menopause index SMI can be used. The Simplified Menopausal Index SMI is obtained by asking each subject to complete the corresponding items in advance according to the degree of the various listed symptoms, and scoring the results. The Simplified Menopausal Index SMI is said to be an index that reflects symptoms specific to Japanese menopausal women. Therefore, depending on the country and regions, other indicators regarding menopause, for example, menopause rating scales (THE MENOPAUSE RATING Scale), Cuperman's menopausal index, and menopausal evaluation scale (GREE) N Climacteric Scale), psst (The premenstrual symptoms screening tool), MRS (Menopause rating Scale), WHQ (The Women's Health Questionnaire), VAS (Visual analysis) scale), HFRDI (Hot Flash Related Daily Interference Scale), HFCS (Hot Flash Composite Score), and It is also possible to use one or more of the Menopause-Specific Quality of Life (MENQOL) and the Menopausal Symptom Evaluation Sheet for Japanese Women as appropriate.

Regarding the health status of women, in addition to menopausal symptoms, premenstrual syndrome (PMS, PMDD) symptoms are also considered. That is, in order to predict premenstrual syndrome, the health prediction model 110Z may be constructed by machine learning using an index indicating the degree of symptoms of premenstrual syndrome as an objective variable and purchasing data 952 as an explanatory variable.

Further, the health data 954Z may include numerical data of these indicators, or may include only labels indicating the degree of symptoms based on the numerical values of these indicators. Regarding this point, the same applies to the following indicators of sleep status, lifestyle-related diseases, etc.

3.2.2 Sleep status (disturbed sleep)
An indicator indicating sleep may be an indicator indicating either deterioration of sleep state and/or sleep rhythm, or inappropriate sleep time. The subjective symptoms of each control obtained through a questionnaire or the like may be used, or the measured values obtained as a result of measuring each control may be used. Sleep status was assessed using the Athens Sleep Scale (AIS), the Pittsburgh Sleep Quality Index (PSQI), the 3 Dimensional Sleep Scale (3DSS), and the Insomnia Severity Questionnaire (Insomnia Severity I). ndex :ISI) can be used. The severity of symptoms is analyzed individually and in combination for each item, and the health status of the consumer is assigned to categories and classified. Sleep rhythm is understood by using the Munich Chronotype Questionnaire (μMCTQ) to find the median sleep time on weekdays and holidays, and calculating the social jet lag (SJL) from the difference. be able to.

3.2.3 Lifestyle-related diseases (lack of physical activity)
The indicator indicating the exercise state may include one or more selected from lack of exercise, obesity, and change in body weight over a certain period of time. More specifically, it is conceivable to use any one or more of each subject's weight, BMI (Body Mass Index), number of steps, and/or metabolic syndrome diagnostic reference value. Metabolic syndrome diagnostic standard values include waist circumference, thresholds established for the diagnosis of hyperglyceridemia and/or low HDL cholesterolemia, maximum and/or minimum blood pressure, and fasting hyperglycemia. It is one or more of the threshold values determined for diagnosis.

In addition, when performing machine learning to generate predictive models related to lifestyle-related diseases and exercise status, in addition to purchase data 952Z, machine learning is performed by including one or more information on the subject's smoking status and sedentary behavior as explanatory variables. You may go.

3.2.4 Moisture status (moisture deficiency degree)
Possible indicators of water status include serum Na value, BUN/creatinine ratio, and/or urine osmolarity, urine specific gravity, urine color, dehydration evaluation scale, and subjective symptoms associated with water deficiency. This information can be obtained by a method similar to the previous example of menopausal symptoms. For example, changes in the body experienced by each of the plurality of controls over a certain period of time, for example, one year, can be obtained from clinical test values obtained during a medical examination or physical checkup, or from a health questionnaire conducted separately. On the other hand, subjective symptoms associated with water loss may be obtained through a questionnaire or the like. The degree of symptoms can be determined from at least one of the following: serum Na value, urine osmolarity, urine specific gravity, BUN/creatinine ratio, water consumption, physical activity amount, and urine color.

Based on these indicators, it is possible to find out from epidemiological information the conditions in which a person is experiencing, or is likely to experience, worsened health conditions in the literature. For example, it has been found that people with serum Na levels over 142 mEq/L have a higher risk of developing cognitive impairment and hypertension than people with serum Na levels less than 142 mEq/L, and are evaluated as having a high health risk. It is possible to do so.

In addition to the purchasing data 952Z, during machine learning to generate a predictive model related to water status, the amount of water consumed by the subject, amount of alcohol consumed, amount of physical activity, urine color, urine osmolarity, urine specific gravity, serum Na level, BUN Machine learning may be performed by including one or more information of /creatinine ratio as an explanatory variable.

3.2.5 Immune status (degree of immunocompromise)
As an index indicating the immune status, for example, ease of catching a cold and/or frequency thereof, and cold-like symptoms can be used. The ease and frequency of catching a cold may be the subjective symptoms of each control obtained through a questionnaire or the like, or may be obtained from the history of visits to medical institutions. The severity of symptoms is aggregated into categories of 3 or more levels, and the severity of symptoms is determined by how easily you catch a cold, SIgA concentration, allergic symptoms, oral environment/immunity status, exercise status, stress status, and severe sleep symptoms. , moderate sleep symptoms, mild sleep symptoms, and cold-like symptoms.

3.2.6 Nutritional status (degree of nutritional deficiency)
The index representing the nutritional status may be a subjective symptom of each control obtained through a questionnaire or the like, or may be a measured value obtained as a result of measuring each control. The severity of the symptoms may be determined by at least one of the following: the presence or absence of subjective symptoms associated with malnutrition, the diversity score (DVS) of food intake, the SNAQ (Simplified Nutritional Appetite Questionnaire), and the BMI (Body Mass Index). .

3.3 Machine Learning Method The model learning program 900Z according to this embodiment uses logistic regression as a machine learning method. However, logistic regression is one example, and other techniques, such as random forests, decision trees, gradient boosting, support vector regression, linear regression, partial least squares (PLS) regression, Gaussian process regression, etc. It is also possible to adopt a neural network or the like. Also, when using logistic regression, accuracy may be further improved by using the L1 regularization method, or logistic regression using the L2 regularization method (ridge regression) may be used instead of the L1 regularization method. You may. It is also possible to use an elastic net that employs both the L1 normalization method and the L2 normalization method.

Logistic regression is a method for determining whether an input corresponds to a certain event by receiving the input and calculating the probability that the input corresponds to the certain event (or the probability that the input does not correspond to the certain event). As an example, a logistic regression model is described as the following equation (1) called a logit function. In equation (1), the left side is the natural logarithm. In addition, i is the i-th subject, p is the probability that the objective variable event occurs, b ₁ , b ₂ ...b _n is the partial regression coefficient, b ₀ is the constant term, x ₁ , x ₂ ... x _n indicates an explanatory variable, respectively.

If the right side of equation (1) is z, equation (1) can be transformed into equation (2) expressing probability p.

式（２）はシグモイド関数と呼ばれ、ロジスティック回帰モデルにおける活性化関数である。ｚの定義における偏回帰係数ｂ_１、ｂ_２…ｂ_ｎは重みと呼ばれ、ｂ_０はバイアス項である。ｚをシグモイド関数に入力すると、０～１の値、すなわち確率値が算出される。

Equation (2) is called a sigmoid function and is an activation function in a logistic regression model. In the definition of z, the partial regression coefficients b ₁ , b _{2 .} . . _bn are called weights, and b ₀ is a bias term. When z is input to the sigmoid function, a value between 0 and 1, that is, a probability value, is calculated.

Here, the sigmoid function obtained by equation (2) is referred to as "φ(z)", and the output y is classified into one of two values according to its output value. Equation (3) means that if φ(z) is 0.5 or more, it is classified into class 1, and if it is less than 0.5, it is classified into class 0. Note that 0.5 is an example, and a value other than 0.5 may be used as the threshold. In other words, equation (3) means that if z in equation (2) is 0 or more, it is classified into class 1, and if z in equation (2) is less than 0, it is classified into class 0. .

Next, we introduce "likelihood" which is used for learning in logistic regression. Likelihood means the likelihood of a condition as seen from the result. A likelihood function L representing the likelihood is expressed by the following equation (4). Note that P represents a probability value.

The likelihood function L indicates the probability of correctly determining all events. By determining the weight that maximizes the likelihood, the probability of the event to be predicted can be output more accurately. Specifically, the likelihood function L is multiplied by (-1) to reverse the government of the likelihood function. This likelihood function whose sign is reversed is the error function in logistic regression. In order to obtain the weights at which the error function has a minimum value, that is, the partial regression coefficients b ₁ , b ₂ . . _{. bn} , partial differentiation and gradient descent are applied to each of the error functions b ₁ , b ₂ . This performs logistic regression learning.

Note that a log likelihood function using the natural logarithm of equation (4) may be used as the likelihood function. The optimal weight can be found by multiplying the log likelihood function by (-1) and minimizing the function with the sign reversed.

The calculation algorithm for the above-mentioned logistic regression is well known, and software applications for performing machine learning using such an algorithm are easily available. Although the principle is as described above, if available software is used, even a person who does not know the details of the specific analysis method using mathematical formulas can operate the machine using the logistic regression according to this embodiment. Learning is possible.

3.4 Method for determining product categories used for health prediction Numerical data such as the purchase amount, purchase quantity, and number of purchases regarding each product and service category in the purchase data 952Z included in the teacher data 950Z is calculated using the above formula (1). It corresponds to the explanatory variable x in etc.

As mentioned above, the health data 954Z included in the teacher data 950Z in this embodiment includes "H" indicating whether each symptom, such as menopausal symptoms or sleep disturbances, is classified into the H layer or the L layer. or "L" (or "H," "M," or "L") labels are assigned to each subject. In Equation (3) above, the likelihood function can be determined by classifying so that "1" is output when the layer is in the H layer, and "0" is output when the layer is other than the H layer. The elements constituting the explanatory variable x are not limited to the above-mentioned purchase data 952Z. For example, explanatory variables may include data indicating age groups and data indicating gender.

By performing machine learning using training data 950Z from a large number of subjects and constructing a predictive model, it is possible to determine product groups (categories) that are often purchased by H-tier customers.

FIG. 24 shows product categories selected (often purchased) by many test subjects in the H and non-H groups regarding menopausal symptoms. The upper part of FIG. 24 shows product categories selected by most subjects in the H group, and the lower part shows product categories selected by subjects other than the H group.

Furthermore, FIG. 25 shows the relationship between menopausal symptoms and health issues found from items purchased by subjects in the H group regarding menopausal symptoms. The numerical values in the rightmost column of FIG. 25 indicate the degree (deviation value) indicating the ease of selection by women with menopausal symptoms. This figure is a quote from ``Symptom-specific deviation values for items that need to be addressed even if it means spending money - Research project report on understanding women's health and beauty lifestyles - Results of a behavioral observation survey.''

From these results, it can be seen that subjects in the H group regarding menopausal symptoms actively select (purchase) specific products. Subjects in the H tier purchase products that address dry skin and internal complaints, as well as indeterminate complaints.

FIG. 26 shows an example of a product group (product category) for women with menopausal symptoms who belong to the H group. In this way, it is possible to determine the product categories that subjects in the H group purchase more frequently compared to subjects in the non-H group. It can be estimated that the subject is likely to belong to the H group in terms of menopausal symptoms. Similarly, it is also possible to determine the product categories that subjects who do not belong to the H group purchase more frequently compared to subjects who are in the H group. It can be estimated that there is a high possibility that it does not belong to the H layer.

Therefore, the model learning program 900Z in this embodiment constructs a prediction model 110Z using data of product categories that can contribute to estimating that the product belongs to the H layer or does not belong to the H layer, out of the purchase data 952Z. This makes it possible to accurately estimate customers belonging to the H group.

3.5 Configuration of Model Learning Program 900Z The functional configuration of the model learning program 900Z will be described below with reference to FIG. 23. FIG. 23 is a diagram showing the functional configuration of the model learning program 900Z. The model learning program 900Z includes an input section 910Z, a product category determination section 912Z, a learning section 914Z, and an output section 916Z. Note that the model learning program 900Z may be realized as a single program, or may be realized as a plurality of programs working together.

Input unit 910Z receives input of teacher data 950Z. The input unit 910Z may read the teacher data 950Z from a storage medium such as an HDD or SSD built into the information processing device (computer) on which the model learning program 900Z is executed, or may read the teacher data 950Z via a network such as a LAN or the Internet. The teacher data 950Z may be input from a connected external information processing device or the like.

The product category determination unit 912Z selects the product category of the purchase data 952Z used for machine learning of the learning target prediction model 110Z. The product category selection method was described in 3.4 above, so detailed explanation will be omitted here. Note that when generating a plurality of predictive models 110Z that respectively correspond to a plurality of symptoms, the product category determination unit 912Z can select different product categories for each of the predictive models 110Z.

Note that, as described above, during machine learning of the prediction model 110Z, it is also possible to create the prediction model 110Z using the entire purchase data 952Z without narrowing down the product category. In that case, the product category determining section 912Z is unnecessary.

The learning unit 914Z performs machine learning using the purchase data 952Z of the product category determined by the product category determining unit 912Z to generate a prediction model 110Z. The method for generating the prediction model 110Z has been described in 3.3 above, so a detailed explanation will be omitted here.

The output unit 916Z outputs the prediction model 110Z generated by the learning unit 914Z to another information processing device (for example, the information processing device 800Z described with reference to FIG. 22) connected via a built-in storage medium or a network. etc. Note that, as an output method, it is possible to output the function itself that constitutes the prediction model 110Z, or to output it as a parameter of the function that constitutes the prediction model 110Z.

3.6 Processing Flow The processing flow of the model learning program 900Z will be described below with reference to FIG. FIG. 27 is a flowchart showing the procedure for creating the predictive model 110Z, which is executed by the model learning program 900Z.

The input unit 910Z of the model learning program 900Z acquires teacher data 950Z regarding a large number of subjects used for learning (S1301). The teacher data 950Z includes purchase data 952Z indicating the subject's purchase history and health data 954Z indicating the health condition (level of symptoms). As described above, the purchase data 952Z can include, for example, JICFS classification (classification code), product category, and corresponding numerical data such as the number of purchases, purchase amount, and purchase quantity.

Next, the product category determination unit 912Z determines the JICFS classification and product category used to generate the prediction model 110Z (S1303). The determination method is as described in 3.4 above, but the product category determination unit 912Z may select the JICFS classification and product category that have a large influence on the estimation of the health condition obtained as the health data 954Z. Note that when the prediction model 110Z is created using the entire purchase data 952Z without particularly narrowing down the product category, the process of S1303 is not necessary.

The learning unit 914Z performs machine learning using the purchase data 952Z as an explanatory variable and the health data 954Z as an objective function (S1305). The machine learning method has been explained with a specific example in 3.3 above, so the explanation will be omitted here.

The output unit 916Z outputs the prediction model 110Z generated by the learning unit 914Z in S1305 to a built-in storage medium, another information processing device connected via a network, or the like. Note that, as described above, the prediction model 110Z may be output as a function itself configuring the prediction model 110Z, or may be output in the form of a parameter or the like.

3.7 Hardware Configuration A specific example of the hardware configuration of the information processing device 1400Z on which the model learning program 900Z is executed will be described with reference to FIG. FIG. 28 is a hardware configuration diagram of an information processing device 1400Z in which a model learning program 900Z is executed. The information processing device 1400Z uses the model learning program 900Z to generate a prediction model 110Z that makes predictions regarding the customer's menopausal symptoms, sleep state, and the like.

The information processing device 1400Z can be realized as a generally available computer system, and can be installed at the office of the business operator operating the health prediction system 1Z or provided as a cloud service, as described above. This is similar to the information processing device 800Z. In this embodiment, the information processing device 800Z that performs health prediction and the information processing device 1400Z that generates the prediction model 110Z are mainly explained as different devices, but it is not possible to implement both as the same device. Good too.

The information processing device 1400Z includes a control section 1410Z, a communication I/F section 1420Z, and a storage device 1430Z.

The control unit 1410Z may include a CPU, ROM, RAM, etc. The control unit 1410Z can execute the model learning program 900Z stored in the storage device 1430Z. Thereby, the information processing device 1400Z is capable of performing various processes related to the construction of the above-mentioned prediction model 110Z in addition to the functions of a general computer. Note that the arithmetic circuit included in the control unit 1410Z does not need to be a CPU, and may be realized by various processors such as an MPU or a GPU. Moreover, it may be realized by a plurality of processors instead of one processor.

The communication I/F unit 1420Z has an input I/F unit 820Z that performs communication in accordance with standards such as an Ethernet (registered trademark) communication terminal, a USB (registered trademark) terminal, IEEE802.11, 4G, or 5G. This can be realized by the communication circuit that performs this. The communication I/F section 1420Z can be connected to a communication network such as an intranet or the Internet, and receives teacher data 950Z prepared by a business operator that operates the learning system 9. Further, the information processing device 1400Z may communicate directly with other devices via the communication I/F unit 1420Z, or may communicate via an access point or the like.

The storage device 1430Z is a storage medium that stores computer programs and data necessary for operating the information processing device 1400Z. The storage device 1430Z may be, for example, an HDD or an SSD. The storage device 1430Z may include a temporary storage element constituted by a RAM such as DRAM or SRAM, and may function as a work area for the control unit 1410Z. The storage device 1430Z stores the model learning program 900Z and the predictive model 110Z generated by the model learning program 900Z (including a table and/or a parameter group that defines the predictive model 110Z).

4 Effects of this Embodiment In the health prediction system 1Z according to this embodiment, it is possible to predict a plurality of health risks of a customer from the customer's purchase data 150Z in the integrated health prediction program 100Z. Based on the individual prediction results, detailed proposals for improving and/or maintaining the current or future health condition can be made for each customer.

5 Additional Notes The configurations of the embodiments described above may be combined or some of the constituent parts may be replaced. Further, the configuration of the present invention is not limited to the above-described embodiments, and various changes may be made without departing from the gist of the present invention.

For example, in the above embodiment, the integrated health prediction program 100Z includes six types of health prediction models, but the inclusion of six types is only an example. It may be 2 or more types and less than 6 types, or it may include 7 or more types of health predictions. Generally speaking, the integrated health prediction program 100Z may include prediction models 110Z for each of the first to Nth types (N is an integer of 2 or more) of health symptoms. The health prediction model 110Z regarding the k-th health symptom of the k-th type (1≦k≦N) inputs the purchase data 150Z regarding the product category to which at least one product belongs, and calculates the degree of the k-th type health symptom. It suffices if it is constructed as a predictive model to be output. This health prediction model 110Z regarding health symptoms of type k (1≦k≦N) uses purchase data 952Z related to the purchase history of each of a plurality of subjects as an explanatory variable, and uses the health status collected from each of those subjects as an explanatory variable. It can be constructed by machine learning using health data 954Z representing the degree of health as an objective variable.

Further, when the degree of health symptoms is expressed numerically, the upper and lower limits of the numerical range defining the class are merely examples, and may be changed as appropriate. The number of defined classes can also be arbitrarily determined by those skilled in the art.

As will be readily understood by those skilled in the art who have the benefit of this disclosure, the various examples described above may be used in appropriate combinations with each other in various patterns, as long as no contradiction arises.

Given the many possible embodiments to which the principles of the invention disclosed herein may be applied, the various illustrated embodiments are merely preferred examples, and the techniques of the claimed invention It should be understood that the scope should not be considered limited to these preferred examples. In fact, the scope of the claimed invention is defined by the appended claims. Therefore, we request that a patent be granted for all inventions that fall within the technical scope of the claimed inventions as inventions of the present inventors.

1 Communication system 2 Communication network (communication line)
10, 10A, 10B server device 20, 20A, 20B terminal device 11, 21 central processing unit (CPU)

Claims (51)

being executed by at least one processor;
Obtain target behavior data that identifies the history of actions performed by the target user,
Target sleep state data that identifies the sleep state of the target user by inputting the target behavior data into an estimation model generated by executing supervised learning, the target sleep state data identifying the sleep state of the target user. Target property data to identify, target rhythm data to identify the sleep rhythm of the target user, target time data to identify the sleep time of the target user, and target category data to identify the sleep category to which the target user belongs. outputting target sleep state data including at least one of the above from the estimation model;
A computer program product, characterized in that it causes the at least one processor to function. Each set of teacher data includes sample behavior data that identifies the history of actions performed by one sample user corresponding to the set, and a sample that identifies the sleep state of one sample user corresponding to the set. obtaining the estimated model generated by inputting a plurality of sets of teacher data including sleep state data into a learning model;
2. The computer program product of claim 1, causing the at least one processor to function as described above. Each set of teacher data includes sample behavior data that identifies the history of actions performed by one sample user corresponding to the set, and a sample that identifies the sleep state of one sample user corresponding to the set. connected via a communication line to the estimation model generated by inputting a plurality of sets of teacher data, including sleep state data, into a learning model;
2. The computer program product of claim 1, causing the at least one processor to function as described above. Each set of teacher data includes sample behavior data that identifies the history of actions performed by one sample user corresponding to the set, and a sample that identifies the sleep state of one sample user corresponding to the set. Obtain multiple sets of teacher data including sleep state data and
Generating the estimated model by inputting the plurality of sets of teacher data to a learning model and causing it to learn;
2. The computer program product of claim 1, causing the at least one processor to function as described above. The acquired plurality of sets of teacher data include a first plurality of sets of teacher data,
Each set of teacher data included in the first plurality of sets of teacher data is
The sample sleep state data includes at least:
sample property data that identifies the sleep property of one sample user corresponding to the group; and sample rhythm data that identifies the sleep rhythm of one sample user corresponding to the group;
Regarding each set of teacher data included in the acquired first plurality of sets of teacher data,
Corresponding to the group based on sample property data identifying the sleep characteristics of one sample user corresponding to the group and sample rhythm data identifying the sleep rhythm of one sample user corresponding to the group. perform preprocessing to generate sample sleep category data that identifies the sleep category to which one sample user belongs;
After the preprocessing, the training data of each set includes the sample behavior data of one sample user corresponding to the set and the sample sleep category data of one sample user corresponding to the set. inputting the first plurality of sets of teacher data to the learning model to generate the estimation model;
outputting target sleep category data identifying a sleep category to which the target user belongs as the target sleep state data from the estimation model;
5. The computer program product of claim 4, causing the at least one processor to function as described above. the plurality of sets of teacher data include a second plurality of sets of teacher data;
Each set of teacher data included in the second plurality of sets of teacher data is
The sample sleep state data includes at least:
the sample property data that identifies the sleep quality of one sample user corresponding to the group and the sample rhythm data that identifies the sleep rhythm of one sample user corresponding to the group; sample sleep category data identifying a sleep category to which one sample user corresponding to the group belongs;
outputting target sleep category data identifying a sleep category to which the target user belongs as the target sleep state data from the estimation model;
A computer program according to any one of claims 2 to 4, which causes the at least one processor to function as follows. The sample quality data identifying the sleep quality of one sample user corresponding to said set may include the performance of said sample user on the Athens Insomnia Scale, the Pittsburgh Sleep Questionnaire, the 3D Sleep Scale, or the Insomnia Severity Questionnaire. Indicates the score calculated based on the answer content,
The sample rhythm data for identifying the sleep rhythm of one sample user corresponding to the group is the absolute value of the difference between the median time of the sample user's weekday sleep time zone and the median time of the holiday sleep time zone. shows the social jet lag, which is
The sleep category to which one sample user corresponding to the group belongs is:
at least one threshold value arranged in correspondence with the score on one of the horizontal axis and the vertical axis; and arranged in correspondence with the social jet lag on the other axis of the horizontal axis and the vertical axis. The computer program according to claim 5 or 6, wherein the computer program is one of a plurality of categories divided by at least one threshold value. the plurality of sets of teacher data include a third plurality of sets of teacher data;
Each set of teacher data included in the third plurality of sets of teacher data is
The sample sleep state data includes at least:
sample property data that identifies the sleep property of one sample user corresponding to the group; and sample rhythm data that identifies the sleep rhythm of one sample user corresponding to the group;
outputting, as the target sleep state data, target property data that identifies the sleep quality of the target user and target rhythm data that identifies the sleep rhythm of the target user from the estimation model;
A computer program according to any one of claims 2 to 4, which causes the at least one processor to function as follows. the plurality of sets of teacher data include a fourth plurality of sets of teacher data;
Each set of teacher data included in the fourth plurality of sets of teacher data is
The sample sleep state data includes at least:
sample property data identifying sleep properties of one sample user corresponding to the set;
Outputting target property data for identifying sleep properties of the target user as the target sleep state data from the estimation model;
A computer program according to any one of claims 2 to 4, which causes the at least one processor to function as follows. the plurality of sets of teacher data include a fifth plurality of sets of teacher data;
Each set of teacher data included in the fifth plurality of sets of teacher data is
The sample sleep state data includes at least:
including sample rhythm data identifying the sleep rhythm of one sample user corresponding to the set;
outputting target rhythm data for identifying the sleep rhythm of the target user as the target sleep state data from the estimation model;
A computer program according to any one of claims 2 to 4, which causes the at least one processor to function as follows. the plurality of sets of teacher data include a sixth plurality of sets of teacher data;
Each set of teacher data included in the sixth plurality of sets of teacher data is
The sample sleep state data includes at least:
sample time data identifying the sleep time of one sample user corresponding to the set;
outputting target time data identifying the sleep time of the target user as the target sleep state data from the estimation model;
A computer program according to any one of claims 2 to 4, which causes the at least one processor to function as follows. determining at least one target product and/or at least one target service that corresponds to the target sleep state data among the multiple products and/or multiple services;
12. A computer program according to any one of claims 1 to 11, causing the at least one processor to function as follows. By inputting the target sleep state data into another estimation model generated by performing supervised learning, the at least one target product and/or the at least one target service can be estimated from the other estimation model. Output identifying proposal data,
13. The computer program product of claim 12, causing the at least one processor to function as described above. By inputting the target sleep state data into a search table that associates the target sleep state data with the plurality of products and/or the plurality of services, the at least one target product and/or the at least one service can be searched from the search table. obtaining proposal data identifying one target service;
13. The computer program product of claim 12, causing the at least one processor to function as described above. 15. A computer program product according to any of claims 1 to 14, wherein the at least one processor includes a central processing unit (CPU), a microprocessor, and/or a graphics processing unit (GPU). comprising at least one processor;
the at least one processor,
Obtain target behavior data that identifies the history of actions performed by the target user,
Target sleep state data that identifies the sleep state of the target user by inputting the target behavior data into an estimation model generated by executing supervised learning, the target sleep state data identifying the sleep state of the target user. Target property data to identify, target rhythm data to identify the sleep rhythm of the target user, target time data to identify the sleep time of the target user, and target category data to identify the sleep category to which the target user belongs. outputting target sleep state data including at least one of the above from the estimation model;
An information processing device configured as follows. The information processing device according to claim 16, which is a terminal device or a server device. The information processing apparatus according to claim 16 or claim 17, wherein the at least one processor includes a central processing unit (CPU), a microprocessor, and/or a graphics processing unit (GPU). A method performed by at least one processor executing computer readable instructions, the method comprising:
The at least one processor executes the instructions, thereby
obtaining target behavior data identifying a history of actions performed by the target user;
Target sleep state data that identifies the sleep state of the target user by inputting the target behavior data into an estimation model generated by executing supervised learning, the target sleep state data identifying the sleep state of the target user. Target property data to identify, target rhythm data to identify the sleep rhythm of the target user, target time data to identify the sleep time of the target user, and target category data to identify the sleep category to which the target user belongs. outputting target sleep state data including at least one of the above from the estimation model;
A method characterized by comprising: 20. The method of claim 19, wherein the at least one processor includes a central processing unit (CPU), a microprocessor, and/or a graphics processing unit (GPU). A method performed by at least one processor executing computer readable instructions, the method comprising:
The at least one processor executes the instructions, thereby
Each set of teacher data includes sample behavior data that identifies the history of actions performed by one sample user corresponding to the set, and a sample that identifies the sleep state of one sample user corresponding to the set. an acquisition step of acquiring a plurality of sets of teacher data, including sleep state data;
By inputting the multiple sets of training data into the learning model and making it learn,
Target sleep state data that identifies the sleep state of the target user by inputting target behavior data that identifies a history of actions performed by the target user, and target properties that identify the sleep quality of the target user. data, target rhythm data that identifies the sleep rhythm of the target user, target time data that identifies the sleep time of the target user, and target category data that identifies the sleep category to which the target user belongs. an estimation model configured to output target sleep state data including
a generation stage that generates
A method characterized by comprising: The obtaining step includes obtaining a first plurality of sets of teacher data included in the plurality of sets of teacher data,
Each set of teacher data included in the first plurality of sets of teacher data is
The sample sleep state data includes at least:
sample property data that identifies the sleep property of one sample user corresponding to the group; and sample rhythm data that identifies the sleep rhythm of one sample user corresponding to the group;
moreover,
The generation step includes:
Regarding each set of teacher data included in the acquired first plurality of sets of teacher data,
Corresponding to the group based on sample property data identifying the sleep characteristics of one sample user corresponding to the group and sample rhythm data identifying the sleep rhythm of one sample user corresponding to the group. performing preprocessing of generating sample sleep category data that identifies a sleep category to which one sample user belongs;
After the preprocessing, the training data of each set includes the sample behavior data of one sample user corresponding to the set and the sample sleep category data of one sample user corresponding to the set. inputting the first plurality of sets of teacher data into the learning model to generate the estimated model;
22. The method of claim 21, comprising: The obtaining step includes obtaining a second plurality of sets of teacher data included in the plurality of sets of teacher data,
Each set of teacher data included in the second plurality of sets of teacher data is
The sample sleep state data includes at least:
the sample property data that identifies the sleep quality of one sample user corresponding to the group and the sample rhythm data that identifies the sleep rhythm of one sample user corresponding to the group; sample sleep category data identifying a sleep category to which one sample user corresponding to the group belongs;
inputting the second plurality of sets of teacher data into the learning model to generate the estimation model;
22. The method of claim 21, comprising: The sample quality data identifying the sleep quality of one sample user corresponding to the set may be the sample user's sleep quality on the Athens Insomnia Scale, the Pittsburgh Sleep Questionnaire, the Three-Dimensional Sleep Scale, or the Insomnia Severity Questionnaire. Indicates the score calculated based on the answer content,
The sample rhythm data for identifying the sleep rhythm of one sample user corresponding to the group is the absolute value of the difference between the median time of the sample user's weekday sleep time zone and the median time of the holiday sleep time zone. shows the social jet lag, which is
The sleep category to which one sample user corresponding to the group belongs is:
at least one threshold value arranged in correspondence with the score on one of the horizontal axis and the vertical axis; and arranged in correspondence with the social jet lag on the other axis of the horizontal axis and the vertical axis. 24. The method according to claim 22 or claim 23, wherein the method is one of a plurality of categories divided by at least one threshold value. an estimation step of inputting target behavior data identifying a history of actions performed by the target user into the estimation model, and outputting target sleep state data identifying the sleep state of the target user from the estimation model;
25. The method of any of claims 21-24, further comprising: 26. A method according to any of claims 21 to 25, wherein the at least one processor comprises a central processing unit (CPU), a microprocessor, and/or a graphics processing unit (GPU). being executed by at least one processor;
Obtain target behavior data that identifies the history of actions performed by the target user,
Target menopausal state data that identifies the state of menopausal symptoms of the target user by inputting the target behavior data into an estimation model generated by executing supervised learning, the target menopausal state data identifying the state of menopausal symptoms of the target user. Target comprehensive state data that identifies a comprehensive state, target first state data that identifies a first symptom state of the target user, and target second state data that identifies a second symptom state of the target user. outputting target menopausal state data including at least one of the following from the estimation model;
A computer program product, characterized in that it causes the at least one processor to function. Each set of teacher data identifies sample behavior data that identifies the history of actions performed by one sample user corresponding to the set, and identifies the state of menopausal symptoms of one sample user corresponding to the set. obtaining the estimated model generated by inputting a plurality of sets of teacher data including sample menopausal state data into a learning model;
28. The computer program product of claim 27, causing the at least one processor to function as described above. Each set of teacher data identifies sample behavior data that identifies the history of actions performed by one sample user corresponding to the set, and identifies the state of menopausal symptoms of one sample user corresponding to the set. connected via a communication line to the estimation model generated by inputting a plurality of sets of teacher data including sample menopausal state data into a learning model;
28. The computer program product of claim 27, causing the at least one processor to function as described above. Each set of teacher data identifies sample behavior data that identifies the history of actions performed by one sample user corresponding to the set, and identifies the state of menopausal symptoms of one sample user corresponding to the set. Obtain multiple sets of teacher data, including sample menopausal status data,
Generating the estimated model by inputting the plurality of sets of teacher data to a learning model and causing it to learn;
28. The computer program product of claim 27, causing the at least one processor to function as described above. The acquired plurality of sets of teacher data include a first plurality of sets of teacher data,
Each set of teacher data included in the first plurality of sets of teacher data is
As the sample menopausal status data, at least:
sample comprehensive state data that identifies the comprehensive state of menopausal symptoms of one sample user corresponding to the group;
Outputting target comprehensive state data that identifies the comprehensive state of menopausal symptoms of the target user as the target menopausal state data from the estimation model;
31. The computer program product of claim 30, causing the at least one processor to function as described above. Each set of teacher data included in the first plurality of sets of teacher data is
As the sample menopausal status data, at least:
including sample overall state data that identifies a score corresponding to the overall state of menopausal symptoms of one sample user corresponding to the set;
Regarding each set of teacher data included in the first plurality of sets of teacher data,
The sample comprehensive state data that identifies a score corresponding to the comprehensive state of menopausal symptoms of one sample user corresponding to the group is converted into the comprehensive state of menopausal symptoms of one sample user corresponding to the group. Perform preprocessing to convert into sample comprehensive state data that identifies the corresponding category,
inputting the first plurality of sets of teacher data after the preprocessing into the learning model to generate the estimation model;
Outputting target comprehensive state data that identifies a category corresponding to the comprehensive state of menopausal symptoms of the target user as the target menopausal state data from the estimation model;
32. The computer program product of claim 31, causing the at least one processor to function as described above. The score corresponding to the overall state of menopausal symptoms of one sample user corresponding to the group is:
Simplified Menopause Index (SMI), Menopause Rating Scale, Kupperman index, Green Climatic Scale, PSST (Premenstrual Symptom) toms Screening Tool), WHQ (Women's Health Questionnaire) , VAS (Visual Analogue Scale), HFRDI (Hot Flash Related Daily Interference Scale), HFCS (Hot Flash Composite Score), MENQOL (Meno Contents of the sample user's responses to the Pause-Specific Quality of Life) or Menopausal Symptom Evaluation Form for Japanese Women 33. The computer program according to claim 32, wherein the score is calculated based on. The acquired plurality of sets of teacher data include a second plurality of sets of teacher data,
Each set of teacher data included in the second plurality of sets of teacher data is
As the sample menopausal status data, at least:
sample first state data identifying a state of a first symptom of one sample user corresponding to the set;
outputting target first state data that identifies the state of the first symptom of the target user as the target menopausal state data from the estimation model;
31. The computer program product of claim 30, causing the at least one processor to function as described above. Each set of teacher data included in the second plurality of sets of teacher data is
As the sample menopausal status data, at least:
sample first state data identifying a score corresponding to a first symptom state of one sample user corresponding to the set;
Regarding each set of teacher data included in the second plurality of sets of teacher data,
Sample first state data that identifies a score corresponding to the state of the first symptom of one sample user corresponding to the set is added to a category corresponding to the state of the first symptom of one sample user corresponding to the set. Perform preprocessing to convert the sample into first state data that identifies the
inputting the second plurality of sets of teacher data after the preprocessing into the learning model to generate the estimation model;
Outputting target first state data that identifies a category corresponding to the first symptom state of the target user as the target menopausal state data from the estimation model;
35. The computer program product of claim 34, causing the at least one processor to function as described above. the acquired plurality of sets of teacher data include a third plurality of sets of teacher data;
Each set of teacher data included in the third plurality of sets of teacher data is
As the sample menopausal status data, at least:
sample second state data identifying a second symptom state of one sample user corresponding to the set;
outputting target first state data that identifies a second symptom state of the target user as the target menopausal state data from the estimation model;
31. The computer program product of claim 30, causing the at least one processor to function as described above. Each set of teacher data included in the third plurality of sets of teacher data is
As the sample menopausal status data, at least:
sample second state data identifying a score corresponding to a second symptom state of one sample user corresponding to the set;
Regarding each set of teacher data included in the third plurality of sets of teacher data,
The sample second state data identifying the score corresponding to the second symptom state of one sample user corresponding to the set is added to the category corresponding to the second symptom state of one sample user corresponding to the set. Perform preprocessing to convert the sample into second state data that identifies the
inputting the third plurality of sets of teacher data after the preprocessing into the learning model to generate the estimation model;
Outputting target second state data that identifies a category corresponding to the second symptom state of the target user as the target menopausal state data from the estimation model;
37. The computer program product of claim 36, causing the at least one processor to function. determining at least one target product and/or at least one target service that corresponds to the target menopausal state data among the multiple products and/or multiple services;
38. A computer program according to any one of claims 27 to 37, causing the at least one processor to function as follows. By inputting the target menopausal state data into another estimation model generated by performing supervised learning, the at least one target product and/or the at least one target service can be estimated from the another estimation model. Output identifying proposal data,
39. The computer program product of claim 38, causing the at least one processor to function as described above. By inputting the target menopausal state data into a search table that associates the target menopausal state data with the plurality of products and/or the plurality of services, the at least one target product and/or the at least one service can be searched from the search table. obtaining proposal data identifying one target service;
39. The computer program product of claim 38, causing the at least one processor to function as described above. 41. A computer program product according to any of claims 27 to 40, wherein the at least one processor comprises a central processing unit (CPU), a microprocessor, and/or a graphics processing unit (GPU). comprising at least one processor;
the at least one processor,
Obtain target behavior data that identifies the history of actions performed by the target user,
Target menopausal state data that identifies the state of menopausal symptoms of the target user by inputting the target behavior data into an estimation model generated by executing supervised learning, the target menopausal state data identifying the state of menopausal symptoms of the target user. Target comprehensive state data that identifies a comprehensive state, target first state data that identifies a first symptom state of the target user, and target second state data that identifies a second symptom state of the target user. outputting target menopausal state data including at least one of the following from the estimation model;
An information processing device configured as follows. The information processing device according to claim 42, which is a terminal device or a server device. 44. The information processing apparatus according to claim 42 or 43, wherein the at least one processor includes a central processing unit (CPU), a microprocessor, and/or a graphics processing unit (GPU). A method performed by at least one processor executing computer readable instructions, the method comprising:
The at least one processor executes the instructions, thereby
obtaining target behavior data identifying a history of actions performed by the target user;
Target menopausal state data that identifies the state of menopausal symptoms of the target user by inputting the target behavior data into an estimation model generated by executing supervised learning, the target menopausal state data identifying the state of menopausal symptoms of the target user. Target comprehensive state data that identifies a comprehensive state, target first state data that identifies a first symptom state of the target user, and target second state data that identifies a second symptom state of the target user. outputting target menopausal state data including at least one of the above from the estimation model;
A method characterized by comprising: 46. The method of claim 45, wherein the at least one processor includes a central processing unit (CPU), a microprocessor, and/or a graphics processing unit (GPU). A method performed by at least one processor executing computer readable instructions, the method comprising:
The at least one processor executes the instructions, thereby
Each set of teacher data identifies sample behavior data that identifies the history of actions performed by one sample user corresponding to the set, and identifies the state of menopausal symptoms of one sample user corresponding to the set. an acquisition step of acquiring a plurality of sets of training data, including sample menopausal status data;
By inputting the multiple sets of training data into the learning model and making it learn,
Target menopausal state data that identifies the state of menopausal symptoms of the target user by inputting target action data that identifies the history of actions performed by the target user, the comprehensive state of the menopausal symptoms of the target user. at least the following: target comprehensive state data that identifies the target user, target first state data that identifies the first symptom state of the target user, and target second state data that identifies the target user's second symptom state. an estimation model configured to output target menopausal state data including one;
a generation stage that generates
A method characterized by comprising: The obtaining step includes obtaining a first plurality of sets of teacher data included in the plurality of sets of teacher data,
Each set of teacher data included in the first plurality of sets of teacher data is
As the sample menopausal status data, at least:
including sample overall state data that identifies a score corresponding to the overall state of menopausal symptoms of one sample user corresponding to the group;
moreover,
The generation step includes:
Regarding each set of teacher data included in the acquired first plurality of sets of teacher data,
The sample comprehensive state data that identifies a score corresponding to the comprehensive state of menopausal symptoms of one sample user corresponding to the group is converted into the comprehensive state of menopausal symptoms of one sample user corresponding to the group. performing a preprocessing step of converting the sample into comprehensive state data that identifies the corresponding category;
After the preprocessing, the training data of each group includes the sample behavior data of one sample user corresponding to the group and the sample comprehensive state data of one sample user corresponding to the group. inputting the first plurality of sets of teacher data into the learning model to generate the estimated model;
48. The method of claim 47, comprising: The obtaining step includes obtaining a first plurality of sets of teacher data included in the plurality of sets of teacher data,
Each set of teacher data included in the second plurality of sets of teacher data is
As the sample menopausal status data, at least:
sample first state data identifying a score corresponding to a first symptom state of one sample user corresponding to the set;
moreover,
The generation step includes:
Regarding each set of teacher data included in the acquired second plurality of sets of teacher data,
Sample first state data that identifies a score corresponding to the state of the first symptom of one sample user corresponding to the set is added to a category corresponding to the state of the first symptom of one sample user corresponding to the set. performing a preprocessing step of converting the sample into first state data identifying the sample;
After the pre-processing, wherein each set of teacher data includes the sample behavior data of one sample user corresponding to the set and the sample first state data of one sample user corresponding to the set. inputting the second plurality of sets of training data into the learning model to generate the estimation model;
48. The method of claim 47, comprising: The acquisition step includes a step of acquiring a third plurality of sets of teacher data included in the plurality of sets of teacher data,
Each set of teaching data included in the third plurality of sets of teaching data is
As the sample menopausal status data, at least:
sample second state data identifying a score corresponding to a second symptom state of one sample user corresponding to the set;
moreover,
The generation step includes:
Regarding each set of teacher data included in the acquired third plurality of sets of teacher data,
The sample second state data identifying the score corresponding to the second symptom state of one sample user corresponding to the set is added to the category corresponding to the second symptom state of one sample user corresponding to the set. performing a preprocessing step of converting the sample into second state data identifying the second state data;
After the preprocessing, wherein each set of teacher data includes the sample behavior data of one sample user corresponding to the set and the sample second state data of one sample user corresponding to the set. inputting the third plurality of sets of teacher data into the learning model to generate the estimated model;
48. The method of claim 47, comprising: 51. The method of any of claims 47 to 50, wherein the at least one processor comprises a central processing unit (CPU), a microprocessor, and/or a graphics processing unit (GPU).