WO2024111652A1

WO2024111652A1 - Living body symptom change predicting system, information processing device, and living body symptom change predicting method

Info

Publication number: WO2024111652A1
Application number: PCT/JP2023/042084
Authority: WO
Inventors: 将貴秦; 淳西野; 哲橋本
Original assignee: ダイキン工業株式会社
Priority date: 2022-11-24
Filing date: 2023-11-22
Publication date: 2024-05-30
Also published as: JP2024076373A

Abstract

The objective of the present invention is to predict a change in a symptom of a living body caused by environmental factors using a smaller amount of data. The present disclosure provides a living body symptom change predicting system 100 comprising an environment sensor 11 and an information processing device 60, wherein: the environment sensor detects environmental factor data relating to environmental factors in a target space; the information processing device includes a control unit for converting the environmental factor data into intermediate data by means of a model that simulates biological characteristics associated with symptoms relating to the environment; and the control unit uses associating information that associates at least the intermediate data and a possibility of a symptom change relating to the environment to predict the possibility of the symptom change from the intermediate data.

Description

System for predicting changes in symptoms of a living body, information processing device, and method for predicting changes in symptoms of a living body

This disclosure relates to a system for predicting changes in symptoms in a living body, an information processing device, and a method for predicting changes in symptoms in a living body.

It is known that environmental factors (such as temperature, humidity, _CO2 , PM2.5, TVOC, formaldehyde, etc.) can cause changes in human symptoms, such as asthma and allergy symptoms. If changes in human symptoms can be predicted to some extent from environmental factor data, it will be possible to prevent patients from avoiding that environment or to control the environment in advance so that the person's symptoms improve.

Technology is known that detects events that may worsen a chronic disease and recommends actions based on the detection results (see, for example, Patent Document 1). Patent Document 1 discloses technology that detects events that may worsen a chronic disease from physiological data and environmental factor data, and recommends favorable actions and medication based on the detection results.

Patent No. 6987042

However, to create correspondence information that can predict changes in symptoms in a living organism from environmental factor data using machine learning or other methods, a huge amount of data is required. This is because it is believed that when a person is affected by the environment, some kind of internal reaction or change occurs, resulting in symptoms, and there is rarely a simple linear relationship between symptoms and the environment.

This disclosure provides technology that predicts changes in biological symptoms caused by environmental factors using less data.

A first aspect of the present disclosure is a method for manufacturing a semiconductor device comprising:
A symptom change prediction system for a living body having an environmental sensor and an information processing device,
The environmental sensor detects environmental factor data related to an environmental factor of a target space;
the information processing device has a control unit that converts the environmental factor data into intermediate data using a model that simulates biological characteristics related to symptoms caused by an environment;
The control unit predicts the possibility of a symptom change from the intermediate data by using correspondence information that associates at least the intermediate data with a possibility of a symptom change depending on an environment.

According to the first aspect of the present disclosure, changes in a person's symptoms caused by environmental factors can be predicted with less data.

A second aspect of the present disclosure is a symptom change prediction system according to the first aspect,
The model that simulates the biological characteristics is the Weber-Fechner law.

A third aspect of the present disclosure is a symptom change prediction system according to the first aspect,
The model that simulates the biological characteristics is a model that outputs different intermediate data depending on the range of values that the environmental factor data takes or whether the environmental factor data satisfies a condition.

A fourth aspect of the present disclosure is a symptom change prediction system according to the first aspect,
The model simulating the biological characteristics is a model that outputs the intermediate data corresponding to the amount of increase or decrease only when the change in the environmental factor data is either an increase or a decrease, or a model that outputs the intermediate data that changes according to the value of the nth derivative.

A fifth aspect of the present disclosure is a symptom change prediction system according to the first aspect,
The model simulating the biological characteristics is a model that outputs the intermediate data corresponding to the integrated value of the environmental factor data, or the intermediate data that differs depending on the history of changes in the past values of the environmental factor data even if the current value of the environmental factor data is the same.

A sixth aspect of the present disclosure is a symptom change prediction system according to the first aspect,
The model that simulates the biological characteristic is a model that outputs the intermediate data according to the duration that the environmental factor data continues to have a value within a predetermined range, or the number of times that the environmental factor data repeats a value within a predetermined range.

A seventh aspect of the present disclosure is a symptom change prediction system according to the first aspect,
The model that simulates the biological characteristics is a model that outputs the intermediate data that changes after a predetermined time has elapsed from the time when the environmental factor data occurs.

An eighth aspect of the present disclosure is a symptom change prediction system according to the first aspect,
The model that simulates the biological characteristics is a log function, an exponential function, or an n-th order function, in which the environmental factor data is input and the intermediate data is output.

A ninth aspect of the present disclosure is a symptom change prediction system according to any one of the first to eighth aspects,
The control unit further predicts the possibility of the symptom change from the intermediate data and the environmental factor data, using correspondence information in which the environmental factor data is associated with each other.

A tenth aspect of the present disclosure is a symptom change prediction system according to any one of the first to ninth aspects,
The possibility of the symptom change is a possibility of worsening or improving any of the following symptoms: allergy symptoms, asthma symptoms, meteorological illness, infectious disease, decreased sleep quality, decreased alertness/drowsiness, autonomic nervous system imbalance, frailty, dementia/delirium, decreased memory, decreased motor function, heat stroke, motion sickness, or VR sickness. An eleventh aspect of the present disclosure is a symptom change prediction system according to any of the first to tenth aspects,
The environmental factor data is a statistical quantity that has been processed by statistical processing.

A twelfth aspect of the present disclosure is a symptom change prediction system according to the ninth aspect,
The control unit predicts the possibility of the symptom change from the actually measured environmental factor data and the intermediate data, or from predicted values of the environmental factor data and the intermediate data.

A thirteenth aspect of the present disclosure is a symptom change prediction system according to any one of the first to twelfth aspects,
The corresponding information is generated using machine learning techniques using information on changes in symptoms reported regarding asthma symptoms, allergy symptoms, weather-related illnesses, infectious diseases, poor sleep quality, decreased alertness/drowsiness, autonomic nervous system imbalance, frailty, dementia/delirium, decreased memory, decreased motor function, heat stroke, motion sickness, or VR sickness as training data.

A fourteenth aspect of the present disclosure is a symptom change prediction system according to the thirteenth aspect,
the control unit generates first correspondence information using a machine learning technique with the intermediate data as an explanatory variable and the symptom change information reported by a plurality of persons as teacher data;
generating second correspondence information using a machine learning technique with the intermediate data as explanatory variables and the symptom change information reported by the individual as teacher data;
The control unit predicts a possibility of a symptom change of an individual based on the possibility of a symptom change predicted by the first correspondence information and the second correspondence information, respectively.

A fifteenth aspect of the present disclosure is a symptom change prediction system according to the thirteenth or fourteenth aspect,
The control unit generates the correspondence information for each season, each target disease, or each allergen in an allergic symptom, and predicts the possibility of a change in symptoms based on the correspondence information generated for each season, each target disease, or each allergen.

A sixteenth aspect of the present disclosure is a symptom change prediction system according to any one of the first to fifteenth aspects,
The control unit displays the intermediate data and the possibility of a symptom change on the same screen.

A seventeenth aspect of the present disclosure is a method for manufacturing a semiconductor device comprising:
An information processing device,
receiving environmental factor data from the environmental sensor relating to an environmental factor of a target space;
a control unit that converts the environmental factor data into intermediate data using a model that simulates biological characteristics related to symptoms caused by the environment;
The control unit predicts the possibility of a symptom change from the intermediate data by using correspondence information that associates at least the intermediate data with a possibility of a symptom change depending on an environment.

According to the seventeenth aspect of the present disclosure, changes in symptoms of a living body caused by environmental factors can be predicted with less data.

An eighteenth aspect of the present disclosure is a method for manufacturing a semiconductor device comprising:
A symptom change prediction method performed by a symptom change prediction system for a living body having an environmental sensor and an information processing device, comprising:
The environmental sensor detects environmental factor data related to an environmental factor of a target space;
A process in which a control unit converts the environmental factor data into intermediate data using a model that simulates biological characteristics related to symptoms caused by the environment;
A process of predicting the possibility of a symptom change from the intermediate data using correspondence information that associates at least the intermediate data with the possibility of a symptom change is performed.

According to the 18th aspect of the present disclosure, changes in symptoms of a living body caused by environmental factors can be predicted with less data.

1 is a diagram outlining an example of a system configuration and a prediction method of a system for predicting changes in a person's symptoms. FIG. FIG. 13 is a diagram showing a modified example of the system configuration of the human symptom change prediction system. FIG. 1 is a diagram illustrating an example of a system configuration of a system for predicting a change in a person's symptom. FIG. 2 illustrates an example of a hardware configuration of an information processing device. 11 is an example of a functional block diagram illustrating functions of an information processing device in a learning phase, divided into blocks. FIG. 4 is a diagram showing an example of environmental factor data acquired by an environmental factor data acquisition unit; FIG. FIG. 13 is a diagram illustrating a learning phase in which a mathematical model is generated from the statistics of environmental factor data, sensory intensity (explanatory variable), and symptom change information (objective variable, teacher data). FIG. 1 is a flowchart illustrating an example of a flow of a learning method using a gradient boosting decision tree. FIG. 1 is an image diagram of a decision tree. 1A and 1B are diagrams illustrating examples of the shapes of a log function, an exponential function, and an n-th order function. FIG. 1 is a diagram showing an example of the shapes of a step function, a sigmoid function, an IF function, and a function having values only in a specific range. FIG. 13 is a diagram showing an example of the shape (tangent) of a first derivative of a quadratic function. 1 is a diagram showing an example of a relationship between time and environmental factor data (e.g., atmospheric pressure) and a relationship between time and intermediate data (e.g., a physiological reaction amount related to meteorological illness). 1 is an example of a diagram showing the relationship between time and environmental factor data (e.g., the amount of pollen dispersed) and the relationship between time and intermediate data (e.g., a physiological response related to hay fever). 1 is a diagram showing an example of a relationship between time and environmental factor data (for example, temperature) and a relationship between time and intermediate data (for example, a physiological reaction amount related to heat stroke). 1 is an example of a graph showing the relationship between time and environmental factor data (e.g., _CO2 concentration) and the relationship between time and intermediate data (e.g., a physiological response amount related to decreased alertness). 1 is an example of a diagram showing the relationship between time and environmental factor data (for example, the amount of allergens) and the relationship between time and intermediate data (the amount of physiological reaction related to allergic symptoms). 1 is a diagram showing an example of a relationship between time and environmental factor data (for example, the amount of house dust) and a relationship between time and intermediate data (for example, the amount of physiological reaction related to an allergic symptom). 1 is an example of a diagram showing the correspondence between environmental factor data (X-axis) and sensory intensity (Y-axis), and a diagram showing the correspondence between environmental factor data and sensory intensity with two straight lines. FIG. 13 is a diagram for explaining how sensory intensity is estimated from environmental factor data and corresponds to a biological response. 11 is an example of a functional block diagram illustrating functions of an information processing device in a prediction phase, divided into blocks. FIG. FIG. 13 is a diagram illustrating a prediction phase in which environmental factor data and sensory intensity are input to an exacerbation risk prediction unit to predict exacerbation risk. FIG. 13 is a diagram comparing the prediction accuracy of asthma exacerbation risk when predicting without and with sensory intensity. FIG. 13 is a diagram showing an example of a map screen of an exacerbation risk displayed on a user terminal. FIG. 13 is a diagram showing an example of a sensation intensity and exacerbation risk screen displayed on a user terminal. FIG. 13 is a diagram showing an example of environmental factor data and symptoms.

Below, as an example of a form for implementing the present disclosure, a system for predicting changes in human symptoms and a method for predicting changes in human symptoms performed by the system for predicting changes in human symptoms will be described.

<Interim data and changes in symptoms in people>
Symptoms that occur in people are influenced by environmental factors. Symptoms include allergy symptoms, asthma symptoms, weather-related illnesses, infectious diseases, poor sleep quality (insomnia, waking up during the night, difficulty falling asleep, difficulty waking up), decreased alertness/drowsiness, autonomic dysfunction, frailty, dementia/delirium, memory loss, motor function loss, heat stroke, motion sickness, VR sickness, etc. In this way, symptoms are not limited to illness.

However, there is no simple linear relationship between the symptoms that people experience and environmental factors. The reason it cannot be considered a simple linear relationship is that environmental factors affect the human body, causing some kind of reaction or change, which results in symptoms. The relationship between environmental factors and symptoms is known to be nonlinear and extremely complex, so creating a model from the relationship between the two using machine learning or other methods requires a huge amount of data.

In this embodiment, intermediate data that indicates the human body's response (symptoms) to environmental factors is used as an explanatory variable, making it possible to accurately predict symptoms with a smaller amount of data. Intermediate data is a numerical value that simulates the response related to the environmental factors that cause symptoms. Such intermediate data can be generated by a model that simulates biological characteristics (described later). The model that creates the intermediate data can itself simulate symptom characteristics. By converting environmental factor data into intermediate data using a model that simulates biological characteristics, it is possible to improve the accuracy of machine learning with a small number of samples.

In this embodiment, this intermediate data may be described using the term "sensation intensity." Also, changes in symptoms may be described using the term "risk of exacerbation."

Specific environmental factors are known to be those related to air quality indoors, in vehicles, or outdoors (current conditions: temperature, humidity, _CO2 , PM2.5, TVOC (total volatile organic compounds), formaldehyde, etc.). Changes in environmental factors may increase the risk of exacerbation of asthma or allergy symptoms. An exacerbation refers to a condition in which a disease such as asthma or allergy symptoms worsens, does not improve with normal treatment, and requires a change in treatment, and exacerbation risk refers to the risk (possibility) of such a condition occurring.

For example, it is known that there is a correlation between asthma and environmental factors. Asthma is a condition whose onset is said to be contributed to by a wide variety of environmental factors. Environmental factors that are known to aggravate asthma are diverse and include PM2.5, VOCs, SOx, NOx, fungi, dust mites, dust, strong odors/fumes, smoke, cold, dry air, and hot air.

The Ministry of Health, Labour and Welfare has published standard and recommended values for favorable environmental factors, not just for suppressing asthma.

Temperature (17℃ to 28℃), humidity (40% to 70%), carbon dioxide concentration (1000ppm or less), formaldehyde (0.08ppm or less)
However, these standard and recommended values are merely guidelines for whether or not there is a risk to human health, and there are multiple factors that contribute to symptoms.

Accordingly, experts say that currently there is no universal solution to asthma or allergy symptoms, and identifying the cause is difficult. There are also patients who do not experience symptoms during consultations or hospitalization for tests, but only experience symptoms at home. It is also said that self-medication at home is important.

In light of these realities, environmental factors that reduce asthma or allergy symptoms cannot be fully addressed by setting standard values for each factor. It is also important that patients themselves are able to approach their indoor environment, and ease of use of environmental control is also considered important.

<Outline of the method for predicting the risk of progression>
1 is a diagram illustrating an outline of a system configuration and a prediction method of an example of a human symptom change prediction system 100. An environmental device 10, an environmental sensor 11, a user terminal 70, and an information processing device 60 are communicatively connected via a network N.

The environmental sensor 11 is installed in preferably each room of a building in which a patient 9 with asthma or allergy symptoms resides. The environmental sensor 11 detects environmental factor data related to the environmental factors of the target space 7 in which the patient 9 resides. Only one environmental sensor 11 may be installed in a building. The environmental device 10 is a device that controls the environment related to air quality, such as an air conditioner, a ventilation device, or an air purifier. The environmental device 10 may have multiple functions, and there may be an environmental device 10 for each function.

The user terminal 70 is a terminal device that displays a map screen, which will be described later. A web browser or a native application is executed on the user terminal 70, and information to be displayed on the display is received from the information processing device 60 via the network N. The patient 9 can view the map screen or the like to understand the environmental conditions that reduce or do not increase the risk of exacerbation. The user terminal 70 can be carried by the patient 9, and does not need to be installed in the same space as the space in which the environmental sensor 11 is installed.

The information processing device 60 is, for example, a server device that processes various information, provides services, and stores files. The information processing device 60 generates a mathematical model, which will be described later, and predicts the risk of exacerbation by inputting sensory intensity and environmental factor data into this mathematical model. Sensory intensity is a variable that indicates the magnitude of the impact that a person will have when a change in the air quality environment becomes a stimulus.

(1) The environmental sensor 11 transmits environmental factor data to the information processing device 60. In the learning phase for generating a mathematical model, the patient 9 also self-reports symptom change information (e.g., worsening, improvement, remission, improvement) to the information processing device 60 via the user terminal 70.

(2) The information processing device 60 inputs the sensory intensity and environmental factor data into a mathematical model that has been generated to predict the risk of exacerbation. This mathematical model uses sensory intensity without requiring vital data, as described below. The mathematical model is correspondence information that associates sensory intensity with the risk of exacerbation of asthma or allergic symptoms.

(3) The information processing device 60 transmits to the user terminal 70 a map screen in which the risk of deterioration is plotted on a map for each environmental factor data.

(4) The user terminal 70 displays a map screen that indicates which environmental settings will reduce or increase the risk of progression. When the user specifies the displayed environmental settings, the user terminal 70 transmits the environmental settings to the information processing device 60.

(5) The information processing device 60 converts the environmental settings into setting information for the environmental device 10 and transmits it to the environmental device 10. This allows the environmental device 10 to control the air quality so as to reduce or prevent an increase in the risk of deterioration.

In this way, in this disclosure, vital data is not used when predicting the risk of exacerbation. Specifically, in addition to environmental factor data, "intermediate data (sensory intensity)" is introduced as a variable that represents the magnitude of the impact that changes in the air quality environment will have on humans (can replace vital data), making it possible to predict the risk of exacerbation with high accuracy. In addition, by converting environmental factor data into intermediate data using a model that simulates biological characteristics, the accuracy of machine learning can be improved with a small number of samples. This makes it possible for patient 9 to avoid that environment, or for environmental control to be performed to create an indoor environment that reduces the risk of exacerbation.

<<Modifications of system configuration>>
The environmental sensor 11 does not have to exist independently, but may be built into the indoor unit 10b as shown in Fig. 2. Fig. 2 shows a modified example of the system configuration of the human symptom change prediction system 100. The human symptom change prediction system 100 mainly has one outdoor unit 10a as a heat source unit, one or more indoor units 10b as utilization units, and a remote control device (hereinafter referred to as "remote control 12") as an input device for inputting commands related to various settings.

The outdoor unit 10a and the indoor unit 10b are called an air conditioner. The outdoor unit 10a and the indoor unit 10b are connected by a refrigerant connection pipe (gas connection pipe GP) to form a refrigerant circuit. In addition, the human symptom change prediction system 100 has multiple communication networks (network NW1, network NW2) that function as transmission paths for signals between each unit. Network NW2 may be wired or wireless.

The remote control 12 is a user interface that accepts settings such as temperature and humidity. The functions of the information processing device 60 may be the same as those in FIG. 1.

(1) The indoor unit 10b has a built-in environmental sensor 11. The indoor unit 10b transmits environmental factor data to the outdoor unit 10a.

(2) The outdoor unit 10a transmits the environmental factor data to the information processing device 60.

(3) The information processing device 60 inputs the sensory intensity and environmental factor data into the mathematical model that was generated to predict the risk of exacerbation.

(4) The information processing device 60 transmits a map screen showing the risk of deterioration for the environmental factor data on a map to the remote control 12 via the outdoor unit 10a and the indoor unit 10b.

(5) The remote control 12 displays a map screen that indicates which environmental settings will reduce or increase the risk of aggravation. When the user specifies the displayed environmental settings, the remote control 12 transmits the environmental settings to the indoor unit 10b.

(6) The indoor unit 10b converts the environmental settings into setting information for the air conditioner and controls the unit itself. This allows the environmental equipment 10 to control the air quality so as to reduce the risk of deterioration or to prevent an increase.

<System configuration of the system for predicting changes in human symptoms>
Next, a system configuration of the human symptom change prediction system 100 will be described with reference to Fig. 3. Fig. 3 is a diagram showing an example of the system configuration of the human symptom change prediction system 100.

The human symptom change prediction system 100 provides various services utilizing IoT to everyone from administrators to general users by having various devices 30, such as air conditioners and lighting, and a cloud-side information processing device 60 communicate via a network N. The edge device 80, devices 30, sensor switches 53, and user terminal 70 are installed on the customer side, and the information processing device 60 is installed in a data center or on the Internet or other cloud. Note that the edge device 80 is a device that centrally manages the devices 30 and sensor switches 53, so the edge device 80 is not necessary.

The equipment 30 refers to any device that consumes power, such as the environmental equipment 10, security equipment, heat source equipment, fire alarms, AHUs (air handling units), electricity meters, and lighting. The sensor switches 53 are various sensors including the environmental sensor 11, lamps, relays, and the like. The equipment 30 and the sensor switches 53 are communicatively connected to the edge device 80 via a dedicated cable or a network such as a LAN. The equipment 30 and the sensor switches 53 may also be communicatively connected to the edge device 80 via wireless communication.

The equipment 30 and the sensor switches 53 are controlled by the edge device 80. In other words, the edge device 80 applies the required operations to the equipment 30 and the sensor switches 53 to suit the purpose of the equipment 30 and the sensor switches 53. The content of the control varies depending on the type of equipment 30 and the sensor switches 53, but for example, if the equipment 30 is an air conditioner, it may include all control related to the functions of the air conditioner, such as the cooling/heating mode, set temperature, air volume, humidity, air direction, etc., which can generally be set on an air conditioner. Control also includes operating modes such as a mode dedicated to pre-season inspection, microcomputer reset, operation stop, and function substitution.

The equipment 30 collects operating data appropriate for the equipment 30 and transmits it mainly periodically to the edge device 80. Periodically means, for example, once per minute, once per 10 minutes, once per 60 minutes, etc., but this can be set by the user or the information processing device 60. Furthermore, the equipment 30 can transmit operating data to the edge device 80 upon request from the edge device 80 or the user terminal 70. The operating data varies depending on the equipment 30, but in the case of an air conditioner, for example, it can include high pressure and low pressure of the refrigerant, refrigerant temperature, fan rotation speed, and microcomputer CPU temperature.

If the device 30 detects an abnormality, it transmits an abnormality code to the edge device 80. The device 30 that detects the abnormality stops operation. The edge device 80 transmits the abnormality code to the information processing device 60. The edge device 80 may perform the same processing for the sensor switches 53. The sensor switches 53 transmit various detection information, such as detected environmental factor data, and transmit abnormality codes to the edge device 80, for example, periodically.

The edge device 80 is a controller that controls the device 30 and the sensor switches 53. If the edge device 80 is not present, the information processing device 60 controls the device 30 and the sensor switches 53. The edge device 80 has the functions of a control device that controls the device 30 and the sensor switches 53, an information processing device that processes operation data, etc., and a communication device that communicates with the information processing device 60. The edge device 80 transmits, for example, an error code received from the device 30 to the information processing device 60, and receives an instruction corresponding to the error code from the information processing device 60. Alternatively, the edge device 80 receives an instruction from the information processing device 60 without transmitting any information to the information processing device 60 (for example, when an instruction is given to the information processing device 60 from the user terminal 70). The edge device 80 converts the instruction into an appropriate instruction according to the model of the device 30 and the sensor switches 53, and transmits it to the device 30 and the sensor switches 53.

The information processing device 60 may be one or more server devices. Although one information processing device 60 is shown in FIG. 3, the information processing device 60 may be installed in several separate units according to function. Furthermore, the functions of the information processing device 60 may be consolidated into one server device. Furthermore, multiple information processing devices 60 with the same functions may be provided, and the multiple information processing devices 60 may communicate with each other and perform processing like a server cluster.

The information processing device 60 inputs the sensory intensity and environmental factor data into a mathematical model to predict the risk of aggravation, and provides it to the user terminal 70, etc. The information processing device 60 can predict the risk of aggravation not only for facilities such as individual homes and buildings, but also for each region, and may share the risk of aggravation with public broadcasting, etc. The information processing device 60 can also provide the user terminal 70 with environmental settings that are favorable for the risk of aggravation, using comprehensively prepared environmental factor data and the risk of aggravation. The information processing device 60 can also transmit instructions to the edge device 80 for the equipment 30 according to the schedule and operations set by the user terminal 70.

Note that although not shown in FIG. 3, an information processing device that generates the mathematical model may exist separately from the information processing device 60. In this case, the mathematical model generated by the information processing device is introduced into the information processing device 60. In this disclosure, for convenience of explanation, it is assumed that the information processing device 60 generates the mathematical model.

The information processing device 60 also has the functionality of a web server. In response to requests from client software (web client) such as a web browser operated by the user, the web server provides the client with screen information written in HTML files, XML, CSS files, JavaScript (registered trademark), etc. An application that uses web mechanisms in this way is called a web app.

It is preferable that the information processing device 60 is compatible with cloud computing. Cloud computing refers to a form of usage in which resources on a network are used without being aware of specific hardware resources. Cloud computing provides users with data and software that users have traditionally used on their local computers as a service via a network. Users can use a variety of services from any terminal by providing a web browser that runs on a personal computer or mobile information terminal, and an Internet connection environment.

The user terminal 70 is a client terminal that displays various screens provided by the information processing device 60. The user terminal 70 may be used by an administrator or a general user (patient 9 in this disclosure). If the device 30 is in a general household, the administrator may be a family member, or the patient 9 may also serve as the administrator. If the device 30 is in a building managed by a company, the administrator may be, for example, a facility manager, a caregiver, or a nurse.

The user terminal 70 can display a wide variety of screens, but some examples include the map screen mentioned above, a list screen of the devices 30 and sensor switches 53 present on the customer's side, an internal company map showing the locations where the devices 30 and sensor switches 53 are located, and an operation screen for operating the devices 30 and sensor switches 53.

The user terminal 70 may be, for example, a PC (Personal Computer), a smartphone, a tablet terminal, a PDA (Personal Digital Assistant), a wearable PC (sunglasses type, wristwatch type, etc.), etc. However, it is sufficient that it has a communication function and can run a web browser. Also, instead of a web browser, the user terminal 70 may run a native app dedicated to the human symptom change prediction system 100.

<Hardware configuration of information processing device>
The hardware configuration of the information processing device 60 will be described with reference to Fig. 4. Fig. 4 is a diagram showing an example of the hardware configuration of the information processing device 60. As shown in Fig. 4, the information processing device 60 has a processor 221, a memory 222, an auxiliary storage device 223, an I/F (Interface) device 224, a communication device 225, and a drive device 226. The hardware components of the information processing device 60 are connected to each other via a bus 207.

The processor 221 has various computing devices such as a CPU (Central Processing Unit). The processor 221 reads various programs onto the memory 222 and executes them. The processor 211 corresponds to the control unit 110 which controls and performs calculations on the entire information processing device 60. In addition to overall control, the control unit 110 performs processing to estimate the sensory intensity related to human senses. The processing to estimate sensory intensity will be described later.

The memory 222 has a primary storage device such as a ROM (Read Only Memory) and a RAM (Random Access Memory). The processor 221 and the memory 222 form a so-called computer, and the processor 221 executes various programs read onto the memory 222.

The auxiliary storage device 223 stores various programs and various data used when the programs are executed by the processor 221.

The I/F device 224 is a connection device that connects the display device 230, which is an example of an external device, and the operation device 240 to the information processing device 60. The display device 230 displays the internal state of the information processing device 60. The operation device 240 is used when the administrator of the information processing device 60 inputs various instructions to the information processing device 60.

The communication device 225 is a communication device for communicating with the edge device 80 and the user terminal 70 via the network N.

The drive unit 226 is a device for setting the recording medium 250. The recording medium 250 here includes media that record information optically, electrically, or magnetically, such as CD-ROMs, flexible disks, and magneto-optical disks. The recording medium 250 may also include semiconductor memory that records information electrically, such as ROMs and flash memories.

The various programs to be installed in the auxiliary storage device 223 are installed, for example, by setting the distributed recording medium 250 in the drive device 226 and reading the various programs recorded on the recording medium 250 by the drive device 226. Alternatively, the various programs to be installed in the auxiliary storage device 223 may be installed by downloading them from the network N via the communication device 225.

<About the function>
Next, the functional configuration of each device of the human symptom change prediction system 100 will be described in detail with reference to Fig. 5. Fig. 5 is an example of a functional block diagram illustrating the functions of the information processing device 60 in the learning phase by dividing them into blocks.

The information processing device 60 has an environmental factor data acquisition unit 61, a symptom change information acceptance unit 62, a statistical quantity calculation unit 63, a sensory intensity estimation unit 64, and a mathematical model generation unit 65. Each of these units of the information processing device 60 is a function or means realized by the control unit 110 of the information processing device 60 executing the commands of a program expanded in the memory 222.

The environmental factor data acquisition unit 61 acquires environmental factor data that indicates the concentration of an environmental factor actually detected by the environmental sensor 11. In this disclosure, the environmental factor data includes temperature, humidity, _CO2 concentration, PM2.5 concentration, TVOC concentration, and formaldehyde concentration, but these are merely examples. For example, the environmental factor data may include pollen and dust.

The environmental factor data acquisition unit 61 may simply receive the environmental factor data periodically transmitted by the environmental sensor 11, or may request the environmental sensor 11 to measure the environmental factor data and receive the environmental factor data in response. It is preferable that the data be acquired at least once a day, and for example, it may be possible to acquire the data at a frequency of once every few minutes to once every few hours.

The symptom change information receiving unit 62 receives symptom change information input to the user terminal 70 by the patient 9 at risk of exacerbation. The symptom change information is information that indicates the presence or absence of symptoms such as asthma or allergy symptoms, and the intensity of the symptoms. For example, the patient 9 inputs "1" if the symptom occurs and "0" if the symptom does not occur as symptom change information for that day every day. The symptom change information may be a value from 0 to 100, not "1" or "0", or may be a 3 to 5 level of intensity. In this way, the symptom change information indicates not only worsening but also improvement. These values correspond to symptoms that have worsened, improved, gone into remission, or improved. When the patient 9 logs into the information processing device 60, the symptom change information receiving unit 62 obtains the patient ID (the patient can be identified). Basic information about the patient 9 is registered in advance in the information processing device 60. For example, the basic information is age, sex, height, allergens (mold, dust, pollen, etc.), underlying diseases, etc. An allergen is an antigen that specifically reacts with the antibodies of a person with asthma or an allergic disease. Generally speaking, an allergen is a substance that causes allergic symptoms.

The statistics calculation unit 63 calculates the statistics of the environmental factor data by processing the environmental factor data through statistical processing. The statistics may be, for example, a maximum value per day, a minimum value per day, an average value per day, or a standard deviation for each environmental factor data. Furthermore, the statistics are not limited to these, and may be a median, an integrated value, or a moving average over the past few hours to several days. The statistics are calculated in order to reduce the processing load of the environmental factor data, so the statistics calculation unit 63 may not be required.

The sensory intensity estimation unit 64 generates intermediate data based on a model that simulates biological characteristics from the statistics of the environmental factor data. As an example, in the present disclosure, the sensory intensity estimation unit 64 estimates sensory intensity from the statistics of the environmental factor data. Details of sensory intensity are described in Figures 19 and 20.

The mathematical model generation unit 65 uses learning data in which the statistical quantities and sensory intensity of environmental factor data are explanatory variables and symptom change information is the objective variable (teacher data) to generate a mathematical model that predicts the risk of exacerbation from the statistical quantities and sensory intensity of environmental factor data. A mathematical model is a simplification of a real object, and a mathematical representation of the relationship between various quantities in accordance with physical laws. However, it is not required that the mathematical model be strict, and it is sufficient that the mathematical model can predict the risk of exacerbation from environmental factor data and sensory intensity.

<Examples of environmental factor data>
The environmental factor data that can be detected by the environmental sensor 11 will be described with reference to Fig. 6. Fig. 6 shows the environmental factor data acquired by the environmental factor data acquisition unit 61. For ease of explanation, Fig. 6 shows the statistics of the environmental factor data. As shown in Fig. 6, the temperature statistics (daily average value, standard deviation, daily maximum value, daily minimum value), humidity statistics, (omitted), PM2.5 concentration statistics, etc. are acquired. The environmental factor data in Fig. 6 is daily statistics, but statistics for a shorter period of time may also be used.

In addition, FIG. 6 shows symptom change information corresponding to the environmental factor data. The symptom change information was entered by patient 9. This symptom change information is either worsened (1) or not worsened (0).

<Mathematical model generation>
Next, a method for generating a mathematical model will be described with reference to Figures 7 to 9. Figure 7 is a diagram for explaining the learning phase in which a mathematical model is generated from the statistics of environmental factor data, sensory intensity (explanatory variable), and symptom change information (target variable, teacher data). The functional configurations of Figures 5 and 7 are the same, but in Figure 7, the blocks are arranged according to the processing flow.

S1: First, the environmental factor data acquisition unit 61 acquires the environmental factor data actually detected by the environmental sensor 11. The environmental factor data is not limited to actual measured values, but may be forecast values. These forecast values may be provided by the Japan Meteorological Agency or may be predicted by the environmental factor data acquisition unit 61 from the actual measured values.

S2: The symptom change information receiving unit 62 also receives symptom change information input by a patient 9 with asthma or allergy symptoms from a user terminal 70 or the like.

S3: Next, the statistics calculation unit 63 calculates the statistics of the environmental factor data (e.g., maximum value per day, minimum value per day, average value per day, standard deviation, etc.). With the above steps, the environmental factor data shown in FIG. 6 is obtained.

S4: Next, the sensory intensity estimation unit 64 estimates the sensory intensity from the statistics of the environmental factor data. When a predicted value is used for the environmental factor data, the sensory intensity is also estimated using the predicted value. Details of the sensory intensity are explained in Figures 19 and 20. The sensory intensity P is estimated by formula (1).

P = klog(I/ _I0 ) ……(1)
where P is the sensory intensity (also called the sensory quantity), I is the intensity of the stimulus, _I0 is the intensity of the stimulus at which the sensory intensity becomes 0, and k is a constant specific to the stimulus. Details of these will be described later.

S5: The mathematical model generation unit 65 then constructs a mathematical model that outputs the risk of exacerbation from the statistical quantities and sensory intensity of the environmental factor data, with the statistical quantities and sensory intensity of the environmental factor data as explanatory variables and the symptom change information as the objective variable (teaching data). If there are six types of environmental factor data, four statistics for each type of environmental factor data, and one sensory intensity for each type of environmental factor data, then 6 x 5 = 30 pieces of data constitute one sample of training data.

In this disclosure, the statistics of the environmental factor data are used for learning, but since the environmental factor data is also included in the sensory intensity, the mathematical model generation unit 65 may learn without using the statistics of the environmental factor data (using sensory intensity).

The process of generating mathematical models is called machine learning. Machine learning is a technology that allows computers to acquire human-like learning capabilities, whereby the computer autonomously generates the algorithms necessary for judgments such as data identification from training data that is previously loaded, and applies these to new data to make predictions. Methods using machine learning may be any of the following methods: supervised learning, unsupervised learning, semi-supervised learning, reinforcement learning, or deep learning, or may be a combination of these learning methods; any learning method for machine learning is acceptable.

There are various algorithms for generating mathematical models using machine learning, but in this disclosure, we will explain, for example, gradient boosting decision trees. Gradient boosting decision trees are a type of supervised learning that combines "gradient descent," "Boosting (ensemble)," and "decision trees."

Figure 8 is a flow chart that explains the flow of the learning method using gradient boosting decision trees. Figure 9 shows an image of the decision tree.

The mathematical model generation unit 65 calculates an initial value from multiple pieces of symptom change information (teacher data) prepared as learning data (S11). The initial value is, for example, an average value. For the sake of explanation, this average value is referred to as predicted value 1.

Next, the mathematical model generation unit 65 calculates, for each piece of learning data, the value obtained by subtracting the average value from the symptom change information as error 1 (S12). Error 1 is positive if the symptom change information is greater than the average, and negative if it is less than the average. The reason why the "value obtained by subtracting the average value from the symptom change information" is used as the error here is because the squared error between the symptom change information and the predicted value is used as the error calculation method, and the gradient obtained by differentiating the squared error is used to calculate the error. Absolute value error may also be used as the error calculation method.

Next, the mathematical model generation unit 65 constructs a decision tree for the purpose of predicting errors (S13). The number of nodes and layers of the decision tree may be limited to a preset range. This decision tree is a weak classifier (boosting). The top tree in Figure 9 is an image of the decision tree that is created first. Because Figure 9 is an image, the number of nodes, etc. is insufficient compared to the amount of input data, but errors are classified into the leaves (ends of the tree) of such a decision tree.

A brief explanation of decision tree construction will be given.
(i) The mathematical model generation unit 65 calculates the entropy based on the ratio of correct answers to incorrect answers in the training data before classification. The mathematical model generation unit 65 may use the Gini coefficient instead of the entropy.
(ii) The mathematical model generating unit 65 calculates the entropy for each branch based on the ratio of correct answers to incorrect answers for each branch when classified by any one attribute.
(iii) The mathematical model generating unit 65 calculates a weighted average of the entropies of (ii). This weighted average may be calculated by multiplying the number of original data by the ratio of the number of data classified into each branch.
(iv) The mathematical model generating unit 65 adopts the attribute with the largest difference (gain) between the entropies of (i) and (iii) as the attribute of the root.
(v) The structure of the lower part of the root is also determined by processes (i) to (iv).

The mathematical model generation unit 65 then calculates a new predicted value using error 1 (S14). Note that although "error 1" is used here, "error n" increases with repetition. Since error 1 is classified for each sample of training data in the leaves of the decision tree, predicted value 1 can be calculated using error 1 for each sample of training data. If multiple errors are classified into one leaf, the average of the errors contained in the leaf is error 1.

If the new predicted value is predicted value 2, then "predicted value 2 = predicted value 1 + learning rate * error 1". The learning rate is a value smaller than 1, and is a hyperparameter that determines how much error is corrected in one decision tree. The learning rate is set appropriately, for example, to 0.05 to 0.3. By repeatedly performing the process of "finding the error, multiplying it by the learning rate, and adding it", the accuracy is improved little by little.

Next, the mathematical model generation unit 65 calculates an error from the symptom change information, which is the teacher data, and the predicted value 2 (S15). The mathematical model generation unit 65 calculates the error from the predicted value 2 for the symptom change information of all the prepared learning data. The error calculated from the predicted value 2 is called error 2.

The mathematical model generation unit 65 repeats steps S13 to S15 until the error has been calculated a certain number of times or until the error falls below a threshold (S16). As shown in the second and third trees in Figure 9, the mathematical model generation unit 65 classifies the previous error using a new decision tree and calculates predicted

values

3, 4... n using the error and

errors

3, 4... n. Since the error changes as the process is repeated, the structure of the decision tree also changes automatically. The n decision trees created in this way are the mathematical model of the present disclosure. The prediction of exacerbation risk using n decision trees in the prediction phase will be described later.

LightGBM (Light Gradient Boosting Machine), XGBoost (eXtreme Gradient Boosting), Catboost (Category Boosting), etc. are known as various frameworks for realizing gradient boosting decision trees in information processing devices. The mathematical model generation unit 65 may use these frameworks.

In addition, predicting the risk of progression using a gradient boosting decision tree is a so-called regression problem (using continuous values to predict one or more values from another), so the algorithms that can be used are not limited to gradient boosting decision trees. There are many algorithms that can be used for regression, including linear regression (multiple regression), logistic regression, neural networks, Bayesian linear regression, SVM regression, ridge regression, lasso regression, and Poisson regression.

For example, deep learning is an algorithm that predicts XYZ based on input data ABC, and then adjusts the weights between neural networks using backpropagation to reduce the error with the training data.

<Physiological response to the environment based on a model that mimics biological characteristics>
This section provides a detailed explanation of models that simulate biological characteristics and convert environmental factor data into intermediate data. Simulating biological characteristics means estimating the effects of environmental factor data on humans (such as physiological reactions and behavior). One such model, Weber-Fechner's law, will be described in detail later.

The environmental factor data is converted into intermediate data using a model that represents a nonlinear relationship as shown below. In other words, the control unit 110 generates intermediate data that correlates with the amount of physiological response to the environment based on a model that simulates these biological characteristics. All of the following models are nonlinear processes that contain quantitative information.

(i) Models that output intermediate data according to the ratio of values taken by the environmental factor data (log function, exponential function, n-th order function)
This model is suitable for expressing the relationship between the amount of dust and asthma symptoms, etc. The change according to the ratio of values means a relationship in which when the environmental factor data increases at a certain ratio, the value of the intermediate data also increases at a specified ratio. Figures 10(a) to 10(c) show examples of the shapes of a log function 331, an exponential function 332, and an n-th order function 333, which are examples of this model.

(ii) Models that output different intermediate data values depending on the range of values that the environmental factor data takes or whether the environmental factor data meets a condition (step functions, sigmoid functions, IF functions, functions that have values only in a specific range)
This model is suitable for expressing the relationship between temperature and sweating or shivering, or the relationship between temperature and the activity level of immune cells. This model is also suitable for expressing the relationship between temperature and humidity and heat stroke, and the relationship between the amount of dust and coughing or sneezing. Figures 11(a) to 11(d) show examples of the shapes of a step function 334, a sigmoid function 335, an IF function 336, and a function 337 having values only in a specific range, which are examples of this model. The IF function 336 is a function that takes a fixed value when a specified condition is met with respect to the environmental factor data, and takes a different fixed value when the specified condition is not met.

(iii) A model that outputs intermediate data according to the way in which the environmental factor data changes, rather than absolute values (n-th order derivatives (slope, acceleration), functions that have values only for changes in one direction)
This model is suitable for expressing physiological reactions in which a small temperature change is not noticed, or in which a burn caused by touching a high temperature does not heal even if the temperature returns to normal. This model is also suitable for expressing the relationship between atmospheric pressure and meteorological illness, or the relationship between acceleration and car sickness. Figure 12 shows an example of the shape (tangent) of the first derivative 338 of the quadratic function.

Figures 13(a) and (b) are diagrams explaining functions that have values that change in only one direction. Figure 13(a) shows the relationship between time and environmental factor data (e.g. air pressure), while Figure 13(b) shows the relationship between time and intermediate data (e.g. a physiological reaction amount related to meteorological illness). For example, if the environmental factor data is air pressure, the physiological reaction amount changes in value only when the air pressure decreases. In other words, intermediate data corresponding to the amount of increase or decrease is output only when the environmental factor data changes to either an increase or decrease. This model is therefore suitable for expressing the relationship between air pressure and meteorological illness, etc.

(iv) A model that outputs intermediate data according to the integrated value of environmental factor data, or intermediate data that varies depending on the change history (hysteresis) of the past values of environmental factor data even if the current value of the environmental factor data is the same. This model is suitable for expressing the relationship between the amount of pollen intake and the amount of antibodies, or between _CO2 concentration and blood oxygen concentration, etc. This model is also suitable for expressing the relationship between the amount of pollen dispersion and hay fever, or between temperature, humidity and heat stroke, etc.

14(a) and (b) are diagrams for explaining intermediate data according to the integrated value of environmental factor data. FIG. 14(a) shows the relationship between time and environmental factor data (e.g., pollen dispersion amount), and FIG. 14(b) shows the relationship between time and intermediate data (e.g., physiological reaction related to hay fever). For example, if the environmental factor data is the pollen dispersion amount, the intermediate data rises according to the integrated value even if the pollen concentration does not change (time T1), and does not change even if the pollen concentration becomes zero because the integrated value does not change (time T2). Therefore, this model is suitable for expressing the relationship between the amount of pollen intake and the amount of antibodies, or between _CO2 concentration and blood oxygen concentration, etc.

Figures 15(a) and (b) are diagrams explaining intermediate data that differs depending on the history (hysteresis) of past environmental factor measurements even when the environmental factor data is the same. Figure 15(a) shows the relationship between time and environmental factor data (e.g., temperature), while Figure 15(b) shows the relationship between time and intermediate data (e.g., physiological response amount associated with heat stroke). For example, the environmental factor data is temperature, and the intermediate data is physiological response amount associated with heat stroke. Even when the temperature starts to drop (time t1), the physiological response amount does not drop immediately, and even when the temperature drops to the same temperature, the physiological response amount does not drop to the same value (time t2). Therefore, this model is suitable for expressing the relationship between pollen dispersion amount and hay fever, or the relationship between temperature, humidity, and heat stroke, etc.

(v) A model that outputs intermediate data according to the duration that the environmental factor data remains within a certain range or the number of times that the environmental factor data repeats a certain range of values. This model is suitable for expressing habituation to temperature. This model is also suitable for expressing the relationship between the number of times an allergen is ingested and the amount of immune reaction (anaphylaxis), etc.

16(a) and (b) are diagrams for explaining the duration of continuous measurement of environmental factor data and intermediate data corresponding to the sustained value. FIG. 16(a) shows the relationship between time and environmental factor data (e.g., _CO2 concentration), and FIG. 16(b) shows the relationship between time and intermediate data (e.g., physiological response amount associated with decreased arousal). For example, the environmental factor data is _CO2 concentration, and the intermediate data is physiological response amount associated with decreased arousal. The physiological response amount increases (time T1) when _CO2 is at or above a certain concentration for a certain period of time, and decreases (time T2) when CO2 is at or below a certain concentration for a certain period of time. Therefore, this model is suitable for expressing symptoms caused by the duration of continuous exposure to environmental factor data and the sustained value.

Figures 17(a) and (b) are diagrams explaining intermediate data according to the number of repetitions at which environmental factor data is measured. Figure 17(a) shows the relationship between time and environmental factor data (e.g., amount of allergen), and Figure 17(b) shows the relationship between time and intermediate data (amount of physiological response related to allergic symptoms). For example, the environmental factor data is the amount of allergen, and the intermediate data is the amount of physiological response related to allergic symptoms. Even if the amount of allergen is the same, the physiological response is greater the second time than the first. Therefore, this model is suitable for expressing symptoms that arise according to the number of times exposed to environmental factor data.

(vi) A model that outputs intermediate data that changes after a certain time has passed since the occurrence of environmental factor data This model is suitable for expressing the relationship of a delayed runny nose after inhaling pollen, or a delayed immune response after contact with bacteria.

FIGS. 18(a) and (b) are diagrams explaining intermediate data corresponding to measured environmental factor data after time has passed since the environmental factor data was measured. FIG. 18(a) shows the relationship between time and environmental factor data (e.g., the amount of house dust), and FIG. 18(b) shows the relationship between time and intermediate data (e.g., the amount of physiological response associated with allergic symptoms). For example, the environmental factor data is the amount of house dust, and the intermediate data is the amount of physiological response associated with allergic symptoms. The amount of physiological response is roughly proportional to the amount of house dust, but changes with a slight delay (a delay of time T) from the time when the amount of house dust changes. Therefore, this model is suitable for expressing symptoms that occur after time has passed since exposure to environmental factor data.

<Regarding sensory intensity>
Next, the Weber-Fechner law will be described in detail as one model that simulates biological characteristics. The Weber-Fechner law provides sensory intensity as an example of the intermediate data. In other words, the present disclosure introduces "sensory intensity" as a variable that represents the magnitude of the impact that a change in the air quality environment will have on humans as a stimulus. The sensory intensity is estimated from air quality environment data based on the Weber-Fechner law proposed by Weber and Fechner.

The Weber-Fechner law is a law that expresses the strength of stimuli felt by humans in a mathematical formula, and is said to approximately apply to all five senses. Sensory intensity P is estimated from the statistics of environmental factor data as shown in formula (1). I, _I0 , and k in formula (1), which are necessary to estimate sensory intensity from air quality environmental data, are determined as follows:
For "stimulus strength I," statistics from environmental factor data are used.
The average value of the environmental factor data is used for "the intensity of the stimulus at which the intensity of sensation becomes 0 (the intensity of the stimulus at which the sensation begins to be felt) I ₀ ". The reason for using the average value is as follows.

It is known that humans have the tendency to adapt to their surroundings. This adaptation progresses toward the average state of the surrounding environment. Taking this into consideration, in this invention, the average value is used as the intensity _I0 of the stimulus at which the stimulus begins to be felt.
The "stimulus-specific constant k" is a different value for each sensation, but in the present invention, it is used as a weighting coefficient to smooth out differences in the degree of influence caused by differences in the range of values and units that environmental factor data (e.g., temperature, humidity, _CO2 , PM2.5, TVOC, formaldehyde, etc.) can take, and is determined for each environmental factor. The determination method is to create an approximation equation using the environmental factor data as the explanatory variable and the exacerbation risk as the objective variable, and to obtain the constant that minimizes the approximation error.

19(a) and (b) are diagrams showing the correspondence between environmental factor data (X-axis) and sensory intensity (Y-axis). Fig. 19(a) is a graph of the following formula (2).
Y = klogX + α ... (2)
Here, k and α may be appropriately designed constants.

Note that equations (1) and (2) are merely examples, and the correspondence between the stimulus and the sensory intensity may be expressed by, for example, equation (3).
Y = logI ... (3)
I is the statistical value of environmental factor data/daily average value.

Furthermore, the sensory intensity may be calculated using a method other than logarithm. Figure 19(b) shows the correspondence between the stimulus and the sensory intensity using two straight lines. In this way, the sensory intensity is saturated in the region where the stimulus is large rather than in the region where the stimulus is small. The correspondence between the stimulus and the sensory intensity may be expressed using three or more straight lines.

Figure 20 explains how sensory intensity is estimated from environmental factor data and corresponds to a biological response. In Figure 20, the surrounding environment A and the human side (inside the ecosystem) B are shown separately.

Surrounding environment A: People feel stimuli due to changes in the environment, such as changes in temperature, humidity, air quality, and smell.

Human side (inside the ecosystem) B: According to the Weber-Fechner law, the stimulus becomes saturated as it becomes larger. In other words, people perceive the intensity of the sensory sensation, not the magnitude of the stimulus. Sensory intensity then changes form and appears as various biological reactions (changes in heart rate, blood pressure, respiratory rate, etc.). Therefore, it can be seen that sensory intensity is information that can replace vital data.

<Prediction of risk of progression>
Next, a prediction method for predicting the risk of exacerbation from environmental factor data and sensory intensity using the mathematical model generated by the mathematical model generating unit 65 will be described.

<<About the function>>
Fig. 21 is an example of a functional block diagram explaining the functions of the information processing device 60 in the prediction phase by dividing them into blocks. In the explanation of Fig. 21, the differences from Fig. 5 will be mainly explained. The information processing device 60 has an environmental factor data acquisition unit 61, a statistics calculation unit 63, a sensory intensity estimation unit 64, and an exacerbation risk prediction unit 66. Each of these units of the information processing device 60 is a function or means realized by the control unit 110 of the information processing device 60 executing commands of a program deployed in the memory 222.

Among these, the exacerbation risk prediction unit 66 corresponds to the mathematical model. The exacerbation risk prediction unit 66 predicts the possibility of symptom change from the intermediate data using correspondence information that associates the intermediate data with the possibility of symptom change. In the present disclosure, the exacerbation risk prediction unit 66 outputs the exacerbation risk from the environmental factor data (actual measurement, forecast) and the sensory intensity estimated based on the environmental factor data. The possibility of symptom change indicates the degree of probability that the symptom will worsen or improve. Taking the exacerbation risk as an example, when the symptom change information input by the user is 1 (worsened) or 0 (not aggravated), the predicted value of the possibility of symptom change also takes a value in the range of 0 to 1. The closer the predicted value is to 1, the more likely the symptom will change for the worse, and the closer the predicted value is to 0, the more likely the symptom will change for the better. The exacerbation risk unit 66 can predict the degree to which the symptom will change. When a forecast value is used for the environmental factor data, the sensory intensity is also estimated using the forecast value. The exacerbation risk is a value predicted by the exacerbation risk prediction unit 66 from the self-reported symptom change information.

FIG. 22 is a diagram explaining the prediction phase in which environmental factor data and sensory intensity are input to the exacerbation risk prediction unit 66 to predict the risk of exacerbation. Note that the explanation of FIG. 22 will mainly explain the differences from FIG. 7. Steps S1, S3, and S4 may be the same as in FIG. 7.

S6: The exacerbation risk prediction unit 66 takes the statistics of the environmental factor data and the sensory intensity estimated from the statistics as input data and outputs the exacerbation risk.

Prediction using gradient boosting decision trees will now be explained. The exacerbation risk prediction unit 66 inputs input data to all of the decision trees created in the learning phase. There are three decision trees in Figure 9, and the input data for each decision tree is classified into one leaf of the decision tree. An error is stored in each leaf. For example, suppose that errors 1 to 3 are classified into leaves 321 to 323, indicated by dotted circles in Figure 9. The exacerbation risk prediction unit 66 estimates the predicted value of symptom change information (exacerbation risk) by summing up the values obtained by multiplying the errors of these leaves by the learning rate for the average value calculated in the learning phase for all decision trees.

Predicted risk of exacerbation = average + learning rate x error 1 + learning rate x error 2 + learning rate x error 3
Thus, in a gradient boosting decision tree, the final predicted value is the average calculated in the learning phase plus all of the errors 1 through n multiplied by the learning rate.
Final prediction value = average + learning rate x error 1 + learning rate x error 2 + ... + learning rate x error n
When the symptom change information in the training data is worsened (1) or not worsened (0), the predicted value is a value between 0 and 1. The exacerbation risk prediction unit 66 may multiply the predicted value by 100 to convert it into a percentage.

<Improving prediction accuracy by using sensory intensity>
The effect of using the sensory intensity will be described with reference to Fig. 23. Fig. 23 is a diagram comparing the prediction accuracy of the asthma exacerbation risk when it is predicted without using the sensory intensity and when it is predicted with the sensory intensity. The prediction accuracy when it is predicted without using the sensory intensity is 61.1%, whereas the prediction accuracy when it is predicted with the sensory intensity is improved to 73.3%.
<Modification of the learning phase>
It is known that symptom change information varies greatly from person to person. This means that in the learning phase, it is effective for the mathematical model generating unit 65 to generate a mathematical model for each individual. The mathematical model generating unit 65 generates a mathematical model (an example of the first correspondence information) for all patients who report symptom change information, and also generates a mathematical model (an example of the second correspondence information) for each individual.

The exacerbation risk prediction unit 66 predicts the exacerbation risk using a mathematical model for multiple people (e.g., all patients) and also predicts the exacerbation risk using an individual's mathematical model. The exacerbation risk prediction unit 66 provides the individual with one or more of the two predicted exacerbation risks, the higher of the two predicted exacerbation risks, or the average of the two predicted exacerbation risks, making it easier for each individual to grasp the exacerbation risk on that day.

Also, since some allergens, such as pollen, become environmental factors depending on the season, it is effective for the mathematical model generation unit 65 to generate a mathematical model for each season. Also, since allergens differ depending on the individual, it is effective for the mathematical model generation unit 65 to generate a mathematical model for each allergen. The allergens of the patient 9 are registered in the information processing device 60. The mathematical model generation unit 65 groups patients 9 with the same allergen and generates a mathematical model based on the reported symptom change information and environmental factor data. Also, since diseases (underlying diseases) that are likely to develop differ depending on the individual, it is effective for the mathematical model generation unit 65 to generate a mathematical model for each target disease. The disease of the patient 9 is registered in the information processing device 60. The mathematical model generation unit 65 groups patients 9 with the same disease and generates a mathematical model based on the reported symptom change information and environmental factor data.

If sufficient data on environmental factors is available, mathematical models can be created for each user's attributes (e.g., gender and age) and area (e.g., prefecture or city/town/village). This is expected to improve the accuracy of predictions of exacerbation risk.

<Examples of risk of exacerbation>
24 is a display example of an exacerbation risk map screen 300 displayed by the user terminal 70. The map screen 300 mainly has a mode selection field 301 and a map display field 302. The mode selection field 301 has a recommendation button 303, an anti-mold and dust mite button 304, an energy saving button 305, and a decision button 306.

The recommendation button 303 is a button that displays the recommended environmental settings based on a combination of mold and dust mite prevention and energy saving performance.

The Anti-Mold & Mite button 304 is a button that displays environmental settings that suppress mold and mites based on at least one of the mold index or mite index.

The energy saving button 305 displays recommended environmental settings based on energy saving performance.

The decision button 306 is a button that accepts control of the environmental equipment with the environmental settings set by the patient 9.

The map display area 302 displays three maps A to C. This is just one example, and maps A to C may be displayed one at a time. Current environmental conditions 308 to 310 are shown on the three maps A to C, respectively.

Map A shows areas with a high risk of aggravation that correspond to the set temperature and humidity. The areas with a low risk of aggravation are divided into two, and areas with a risk of aggravation below a first threshold are shown in blue (one example) and areas with a risk of aggravation below a second threshold are shown in red (one example) in correspondence with the temperature and humidity of the map screen original data. The map screen original data is data in which environmental factor data and aggravation risk are comprehensively calculated. As the map screen original data is discrete data, the screen generation unit 73 performs processing such as coloring by interpolating between points.

・Map B shows areas with low risk of aggravation associated with _CO2 concentration and PM2.5 concentration. The areas with low risk of aggravation are divided into two, and areas with aggravation risk below the first threshold are shown in blue (example), and areas below the second threshold are shown in red (example), in association with the _CO2 concentration and PM2.5 concentration of the original data on the map screen. However, the first threshold is less than the second threshold.

　Map C shows areas with low risk of aggravation, which are associated with formaldehyde concentration and TVOC concentration. The areas with low risk of aggravation are divided into two, and areas with aggravation risk below a first threshold are shown in blue (one example), and areas below a second threshold are shown in red (one example), which are associated with formaldehyde concentration and TVOC concentration in the original data on the map screen.

This allows users to easily determine what temperature and humidity to set their air conditioner to.

In addition, instead of the user terminal 70 displaying the map screen, the information processing device 60 may display the map screen on the display device 230.

FIG. 25 is an example of a sensory intensity and aggravation risk screen 310 displayed on the user terminal 70. The sensory intensity and aggravation risk screen 310 displays a message 311 saying "The risk of allergic symptoms aggravating is as follows," the aggravation risk 312, and the degree of influence 313 of environmental factors that contribute to the aggravation risk.

The user can check the displayed exacerbation risk 312 and visually grasp the environmental factors that contribute to the exacerbation risk. The sensory intensity and exacerbation risk screen 310 also displays environmental factors 314 that contribute to the exacerbation risk and have a degree of influence equal to or greater than a threshold. This allows the user to know which environmental factors they should adjust.

<Environmental factor data and more detailed description of symptoms>
26(a) and (b) show examples of environmental factor data and symptoms. FIG. 26(a) shows environmental factor data that can be detected by an environmental sensor. In addition to temperature, humidity, pollen, mold, and mites, environmental factor data includes radiation temperature, air pressure, dust (amount, suspended concentration), airflow (wind speed, air volume), CO2, oxygen (concentration), sound (sound pressure, volume, frequency, tempo), odor (intensity), acceleration/inclination (environmental factor data when riding in a vehicle), droplets/steam/smoke (environmental factor data related to infectious diseases), etc. In the present disclosure, these can be source data for predicting the possibility of a change in symptoms.

FIG. 26(b) shows symptoms caused by environmental factor data. In addition to allergies, symptoms include weather-related illnesses, infectious diseases, sleep quality, autonomic nervous system imbalance, dementia/delirium, weakness/frailty, memory decline/improvement, alertness/drowsiness, heat stroke, motor function, motion sickness, and VR sickness. In this disclosure, these can also be symptom changes predicted by the mathematical model. When the mathematical model is being trained, the user reports the presence or absence and severity of these symptoms as training data.

<Major Effects>
As described above, in the present disclosure, vital data is not used to predict the risk of exacerbation. In addition, in the present disclosure, by introducing "sensation intensity" as a variable that represents the magnitude of the impact that a change in the air quality environment will have on a human being, it is possible to replace vital data and predict the risk of exacerbation of symptoms with high accuracy. In addition, by converting environmental factor data into intermediate data using a model that simulates biological characteristics, the accuracy of machine learning can be improved with a small number of samples.

<Other application examples>
The above describes the best mode for carrying out the present disclosure using examples, but the present disclosure is not limited to these examples in any way, and various modifications and substitutions can be made within the scope that does not deviate from the gist of the present disclosure.

For example, the information processing device 60 and the environmental sensor 11 do not have to be separate systems, and the information processing device 60 may be integrated into the environmental sensor 11. In this case, the environmental sensor 11 can display the risk of exacerbation predicted from the environmental factor data of the space in which it is installed on a liquid crystal display or the like.

Similarly, if the indoor unit 10b of the air conditioner has an environmental sensor 11 built in, either the indoor unit 10b, the outdoor unit 10a, or the remote control 12 may display on the remote control 12 the risk of aggravation predicted from the environmental factor data of the space.

In addition, in the present disclosure, vital data is not required to predict the risk of exacerbation, but the information processing device 60 may predict the risk of exacerbation using sensory intensity and vital data.

Furthermore, this embodiment is not limited to predicting symptoms in humans, but may also be used to predict symptoms in animals such as pets, livestock such as cows and pigs, and farmed fish.

Furthermore, the configuration examples in Figures 5, 21, etc. are divided according to main functions to make it easier to understand the processing by the information processing device 60. The technology disclosed herein is not limited by the manner in which the processing units are divided or the names of the processing units. The processing of the information processing device 60 can also be divided into even more processing units depending on the processing content. Also, it can be divided so that one processing unit includes even more processes.

Furthermore, the devices described in the examples are merely representative of one of multiple computing environments for implementing the embodiments disclosed herein. In one embodiment, the information processing device 60 includes multiple computing devices, such as a server cluster. The multiple computing devices are configured to communicate with each other via any type of communication link, including a network, shared memory, etc., and perform the processing disclosed herein.

Each of the functions of the present disclosure described above can be realized not only by software processing through the execution of a program, but also by one or more processing circuits. Here, the "processing circuit" in this specification includes a processor programmed to execute each function by software, such as a processor implemented by an electronic circuit, an ASIC (Application Specific Integrated Circuit), a DSP (Digital Signal Processor), an FPGA (Field Programmable Gate Array), and devices such as conventional circuit modules designed to execute each of the functions described above.

<Reasons for the effect>
The first aspect of the present disclosure is a method for detecting an environmental factor by using an environmental factor analysis method, comprising:
The control unit predicts the possibility of a symptom change from the intermediate data using correspondence information that matches at least the intermediate data with the possibility of a symptom change due to the environment. Therefore, even if there is no simple linear relationship between the symptoms that a person experiences and environmental factors and there is a complex correspondence relationship, it is possible to create correspondence information that matches the intermediate data with the possibility of a symptom change due to the environment with a smaller number of learning data samples.

- The second aspect of the present disclosure is that "the model that mimics the biological characteristics is the Weber-Fechner law," so that the magnitude of the impact that changes in the air quality environment will have on humans can be appropriately converted into intermediate data, and correspondence information that matches the intermediate data with the possibility of changes in symptoms in response to the environment can be created with a smaller number of learning data samples.

- The third aspect of the present disclosure is that "the model simulating biological characteristics outputs different intermediate data depending on the range of values that the environmental factor data takes or whether the environmental factor data satisfies a condition," so that sweating or shivering due to temperature, or the relationship between temperature and the degree of immune cell activity, etc., can be represented by intermediate data, and correspondence information that matches the intermediate data with the possibility of symptom changes due to the environment can be created with a smaller number of learning data samples.

- The fourth aspect of the present disclosure is that "the model simulating biological characteristics outputs the intermediate data according to the amount of increase or decrease only when the change in the environmental factor data changes to either an increase or a decrease, or outputs the intermediate data that changes according to the value of the nth derivative," so that physiological reactions such as not noticing small temperature changes, physiological reactions in which a burn caused by touching a high temperature does not heal even when the temperature returns to normal, the relationship between air pressure and meteorological illnesses, or the relationship between acceleration and car sickness can be represented by intermediate data, and correspondence information that matches the intermediate data with the possibility of changes in symptoms due to the environment can be created with a smaller number of learning data samples.

- In the fifth aspect of the present disclosure, "the model simulating biological characteristics outputs the intermediate data according to the integrated value of the environmental factor data, or the intermediate data which differs depending on the history of changes in the past values of the environmental factor data even if the current value of the environmental factor data is the same", so that relationships between pollen intake and antibody amount, _CO2 concentration and blood oxygen concentration, the relationship between pollen dispersion amount and hay fever, or the relationship between temperature, humidity and heat stroke, etc. can be represented by intermediate data, and correspondence information that matches the intermediate data with the possibility of changes in symptoms due to the environment can be created with a smaller number of learning data samples.

- The sixth aspect of the present disclosure is that "the model simulating biological characteristics outputs the intermediate data according to the duration that the environmental factor data maintains a value within a specified range, or the number of times that the environmental factor data repeats a value within a specified range," so that the relationship between habituation to temperature, the number of times an allergen is ingested, and the amount of immune response (anaphylaxis), etc., can be expressed by the intermediate data, and correspondence information that matches the intermediate data with the possibility of symptom changes in response to the environment can be created with a smaller number of learning data samples.

- The seventh aspect of the present disclosure is that "the model simulating biological characteristics outputs the intermediate data that changes after a predetermined time has elapsed from the time the environmental factor data occurs," so that the relationship of a runny nose occurring with a delay after inhaling pollen, or the relationship of the immune system acting with a delay after coming into contact with bacteria, can be expressed in intermediate data, and correspondence information that associates the intermediate data with the possibility of symptom changes in response to the environment can be created with a smaller number of learning data samples.

- In the eighth aspect of the present disclosure, "the model simulating the biological characteristics is a log function, exponential function, or n-th order function that inputs the environmental factor data and outputs the intermediate data," so that the relationship between the amount of dust and asthma symptoms, etc., can be expressed by intermediate data, and correspondence information that associates the intermediate data with the possibility of symptom changes in response to the environment can be created with a smaller number of training data samples.

- In the ninth aspect of the present disclosure, "the control unit further predicts the possibility of the symptom change from the intermediate data and the environmental factor data using correspondence information that associates the environmental factor data," so that the possibility of a symptom change can be predicted not only from the intermediate data but also from the environmental factor data.

- The tenth aspect of the present disclosure is that "the possibility of symptom change is the possibility of worsening or improving any of the following symptoms: allergy symptoms, asthma symptoms, weather-related illness, infectious disease, decreased sleep quality, decreased alertness/drowsiness, autonomic nervous system imbalance, frailty, dementia/delirium, decreased memory, decreased motor function, heat stroke, motion sickness, or VR sickness," making it possible to predict the possibility of changes in various symptoms.

- In an eleventh aspect of the present disclosure, "environmental factor data is a statistical quantity processed by statistical processing," so that the statistical processing can be converted into intermediate data rather than the environmental factor data itself, and the correspondence information associates this intermediate data with the possibility of symptom changes in response to the environment, thereby improving the accuracy of prediction.

- The twelfth aspect of the present disclosure "predicts the possibility of a symptom change from the actually measured environmental factor data and the intermediate data, or from the forecast values of the environmental factor data and the intermediate data," so that the possibility of a symptom change can be predicted not only from the actually measured environmental factor data and intermediate data, but also from the forecast values of the environmental factor data and the intermediate data.

- A thirteenth aspect of the present disclosure "uses the intermediate data as explanatory variables, and information on changes in symptoms reported regarding asthma symptoms, allergy symptoms, weather-related illnesses, infectious diseases, poor sleep quality, decreased alertness/drowsiness, autonomic dysfunction, frailty, dementia/delirium, poor memory, decreased motor function, heat stroke, motion sickness, or VR sickness as training data, and generates the correspondence information using machine learning techniques," so that correspondence information that learns the correspondence between explanatory variables and training data can be generated, and the correspondence information can predict the possibility of these symptom changes.

- A fourteenth aspect of the present disclosure "predicts the possibility of a symptom change for an individual based on the possibility of symptom change predicted by" first correspondence information for multiple people and second correspondence information for an individual, so that correspondence information for multiple people and for an individual can be generated separately, and the possibility of symptom change is predicted using each of the two pieces of correspondence information, so that information such as those with a high risk of exacerbation can be provided for each individual.

- The fifteenth aspect of the present disclosure "predicts the possibility of a symptom change based on the correspondence information generated for each season, each target disease, or each allergen," making it possible to generate and predict the possibility of a symptom change for each season, each target disease, or each allergen.

- The sixteenth aspect of the present disclosure "displays the intermediate data and the possibility of symptom change on the same screen," making it easy to understand how likely a symptom change is compared to the current intermediate data.

This application claims priority based on Patent Application No. 2022-187732, filed with the Japan Patent Office on November 24, 2022, and the entire contents of Patent Application No. 2022-187732 are incorporated herein by reference.

10 Environmental device 11 Environmental sensor 60 Information processing device 70 User terminal 100 System for predicting changes in human symptoms

Claims

A symptom change prediction system for a living body having an environmental sensor and an information processing device,
The environmental sensor detects environmental factor data related to an environmental factor of a target space;
the information processing device has a control unit that converts the environmental factor data into intermediate data using a model that simulates biological characteristics related to symptoms caused by an environment;
The control unit is a symptom change prediction system for a living body that predicts the possibility of a symptom change from the intermediate data using correspondence information that associates at least the intermediate data with the possibility of a symptom change depending on the environment.
The system for predicting changes in symptoms in a living body according to claim 1, wherein the model that simulates the biological characteristics is the Weber-Fechner law.
The system for predicting changes in symptoms of a living body according to claim 1, wherein the model simulating the biological characteristics is a model that outputs different intermediate data depending on the range of values that the environmental factor data takes or whether the environmental factor data satisfies a condition.
The symptom change prediction system for a living body according to claim 1, wherein the model simulating the biological characteristics is a model that outputs the intermediate data according to the amount of increase or decrease only when the change in the environmental factor data changes to either an increase or a decrease, or a model that outputs the intermediate data that changes according to the value of an n-th order derivative.
The symptom change prediction system for a living body according to claim 1, wherein the model simulating the biological characteristics is a model that outputs the intermediate data according to an integrated value of the environmental factor data, or the intermediate data that differs depending on a history of changes in the past values of the environmental factor data even if the current value of the environmental factor data is the same.
The symptom change prediction system for a living body according to claim 1, wherein the model simulating the biological characteristics is a model that outputs the intermediate data according to the duration for which the environmental factor data maintains a value within a predetermined range, or the number of times that the environmental factor data repeats a value within a predetermined range.
The symptom change prediction system of claim 1, wherein the model simulating the biological characteristics is a model that outputs the intermediate data that changes after a predetermined time has elapsed from the time when the environmental factor data occurs.
The system for predicting changes in symptoms of a living body according to claim 1, wherein the model simulating the biological characteristics is a log function, an exponential function, or an n-th order function, in which the environmental factor data is input and the intermediate data is output.
The system for predicting a symptom change in a living body according to any one of claims 1 to 8, wherein the control unit further predicts the possibility of the symptom change from the intermediate data and the environmental factor data using correspondence information that associates the environmental factor data.
The system for predicting changes in symptoms in a living body according to claim 1, wherein the possibility of a change in symptoms is the possibility of a worsening or improvement in any of the following symptoms: allergy symptoms, asthma symptoms, weather-related illness, infectious disease, poor sleep quality, decreased alertness/drowsiness, autonomic dysfunction, frailty, dementia/delirium, poor memory, poor motor function, heat stroke, motion sickness, or VR sickness.
The system for predicting changes in symptoms in a living body according to claim 1, wherein the environmental factor data is a statistical quantity processed by statistical processing.
The system for predicting symptom changes in a living body according to claim 9, wherein the control unit predicts the possibility of the symptom change from the actually measured environmental factor data and the intermediate data, or from the predicted values of the environmental factor data and the intermediate data.
The control unit classifies the intermediate data into explanatory variables,
The change in symptoms reported for asthma symptoms, allergy symptoms, weather-related illnesses, infectious diseases, poor sleep quality, decreased alertness/drowsiness, autonomic nervous system imbalance, frailty, dementia/delirium, memory loss, motor function loss, heat stroke, motion sickness, or VR sickness is used as training data.
The system for predicting changes in symptoms of a living body according to claim 1 , wherein the corresponding information is generated using a machine learning technique.
the control unit generates first correspondence information using a machine learning technique with the intermediate data as an explanatory variable and the symptom change information reported by a plurality of persons as teacher data;
generating second correspondence information using a machine learning technique with the intermediate data as explanatory variables and the symptom change information reported by the individual as teacher data;
The system for predicting symptom changes in a living body as described in claim 13, wherein the control unit predicts the possibility of a symptom change in an individual based on the possibility of symptom change predicted by the first correspondence information and the second correspondence information, respectively.
The control unit generates the correspondence information for each season, each target disease, or each allergen in an allergic symptom,
15. The system for predicting a symptom change in a living body according to claim 13 or 14, which predicts the possibility of the symptom change based on the correspondence information generated for each season, each target disease, or each allergen.
The system for predicting changes in symptoms in a living body according to any one of claims 1 to 8, wherein the control unit displays the intermediate data and the possibility of the symptom change on the same screen.
An information processing device,
receiving environmental factor data from an environmental sensor relating to an environmental factor of the target space;
a control unit that converts the environmental factor data into intermediate data using a model that simulates biological characteristics related to symptoms caused by the environment;
The control unit is an information processing device that predicts the possibility of a symptom change from the intermediate data by using correspondence information that associates at least the intermediate data with a possibility of a symptom change depending on an environment.
A symptom change prediction method performed by a symptom change prediction system for a living body having an environmental sensor and an information processing device, comprising:
The environmental sensor detects environmental factor data related to an environmental factor of a target space;
A process in which a control unit converts the environmental factor data into intermediate data using a model that simulates biological characteristics related to symptoms caused by the environment;
A process of predicting a possibility of a symptom change from the intermediate data using correspondence information that associates at least the intermediate data with a possibility of a symptom change depending on an environment;
A method for predicting changes in symptoms.