JP7426015B1 - Healthcare data analysis system - Google Patents

Abstract

An object of the present invention is to provide a healthcare data analysis system that can quickly and accurately analyze time-series data of variables related to personal healthcare. [Solution] In a server device 2, after collecting time-series data of variables related to each user's health care, a first relationship between variables related to each user's health care is determined by machine learning by a predetermined analysis process. The predetermined analysis process is changed to match or approximate the second relationship between variables related to the user's health care. Further, in the user information terminal device 1, the predetermined analysis process is changed based on the change information of the predetermined analysis process in the server device 2, and the changed analysis process is performed based on the time series data of variables related to the user's health care. By doing this, the relationships between variables are newly calculated. [Selection diagram] Figure 2

Description

The present invention relates to a healthcare data analysis system that analyzes time-series data of variables related to a user's healthcare.

In recent years, in conjunction with the need for physical condition management to prevent pandemics and the health boom, collecting and visualizing changes in a person's internal state can help maintain health and beauty while being aware of changes in one's own internal state. This is attracting attention.

Particularly recently, various wearable devices such as glasses, wristwatches, and wristbands that are worn on the human body have appeared. Applications are also available that use sensors to collect changes in the state of a person's body and visualize those changes through the device itself or a smartphone.

However, although it has become possible to manage health and beauty information on smartphones by linking with wearable devices, the data that can be managed on smartphones is only information on steps or heart rate. In some cases, information on blood pressure was the only information, so there was only a single piece of information, and there was little comprehensive management.

Therefore, as a system for centrally managing physical activity information that can be obtained from multiple wearable devices, a health management system called "Google Fit (registered trademark)" is available for Android devices and "Healthcare" for iOS devices. There is.

For example, "Google Fit" allows information related to fitness data provided by Google to be collectively stored and referenced across devices and application users. Furthermore, "Google Fit" alone can record activity information, location information data, body measurements, nutritional information, and sleep information using the location information of the Android terminal and various sensors. On the other hand, "Healthcare" can record activity information, vital information, physical measurements, reproductive health information, test results, nutritional information, mindfulness, sleep information, etc. from the device's location information and various sensors, It is possible to read and save information recorded in "Healthcare" from the app.

Furthermore, the present applicant has proposed a system for estimating relationships between variables (attributes) that are significant to people, based on time-series data of a plurality of variables (attributes) related to the health care of individual users.

Biofeedback Research, 2017, Volume 44, No. 2 "Current status, problems, and future prospects of systems that utilize wearable devices"

Japanese Patent Application Publication No. 2022-013409

By the way, in each of the above-mentioned systems, analysis processing is often executed on a local user information terminal device because it handles time-series data of variables related to the user's health care, but in this case, the results of the analysis processing cannot be quickly delivered. However, since the initially set closed analysis process continues to be used for a long period of time, there is a problem that the results of the analysis process lack accuracy. Of course, it is also possible to execute the latest analysis processing using AI (Artificial Intelligence) on a server device on the cloud and return the results of the analysis processing to the user's information terminal device, but in that case, the analysis There was a problem in that the processing results were not quick enough. As described above, the results of the analysis process lack either accuracy or speed, and therefore, although many of the above-mentioned systems have been provided, they have not become widespread at present.

The present invention has been made in view of the above-mentioned problems, and is capable of quickly and accurately analyzing time-series data of variables related to personal health care, thereby contributing to the spread of a system related to user health care. The purpose is to provide a healthcare data analysis system that can perform the following tasks.

In order to achieve the above object, the present invention provides a healthcare data analysis system in which a plurality of user information terminal devices and a server device are connected via a network, and which analyzes time-series data of variables related to the user's healthcare. The user information terminal device includes a data acquisition unit that acquires time series data of variables related to the user's health care, and a data acquisition unit that acquires time series data of variables related to the user's health care, and a predetermined data acquisition unit based on the time series data of the variables related to the user's health care acquired by the data acquisition unit. a user-side analysis processing unit that performs analysis processing; and a user-side relationship that calculates a relationship between variables related to the user's health care based on the results of a predetermined analysis processing performed by the user-side analysis processing unit. a calculation unit; a user-side transmission unit that transmits time-series data of variables related to the user's health care acquired by the data acquisition unit to the server device via a network; and a user-side receiving unit that receives change information regarding a predetermined analysis process, and the server device receives time-series data of variables related to each user's health care transmitted from each user information terminal device via the network. a server-side receiving unit that receives the information, a server-side analysis processing unit that performs a predetermined analysis process based on the time-series data of variables related to healthcare of each user received by the server-side reception unit, and the server-side analysis process. a first server-side relationship calculation unit that calculates a first relationship between variables related to the health care of the entire user based on the results of the analysis process performed by the unit; a learning unit that performs machine learning based on time-series data of variables related to each user's healthcare; and a second relationship between variables related to healthcare of all users based on the results of machine learning performed by the learning unit. The second server-side relationship calculation unit that calculates the relationship of an analysis for changing a predetermined analysis process in the first server-side relationship calculation unit so as to match or approximate a second relationship between variables related to healthcare for all users calculated by the side relationship calculation unit; a processing change unit; and a server-side transmission unit that transmits change information regarding a predetermined analysis process changed by the analysis process change unit to the user information terminal device via a network; The user-side analysis processing unit changes the predetermined analysis process based on the change information regarding the predetermined analysis process received by the user-side reception unit, and then changes the user's health care information acquired by the data acquisition unit. The user-side relationship calculation unit performs a predetermined analysis process after the change based on time-series data of variables related to , is characterized by newly calculating the relationship between variables related to the user's health care.

According to this, in a user information terminal device owned or used by a user, relationships between variables are calculated by performing a predetermined analysis process based on time-series data of variables related to the user's health care. can quickly analyze time-series data on variables related to healthcare. In addition, the server device collects time-series data of variables related to each user's healthcare, and after changing a predetermined analysis process using machine learning, the user information terminal device collects time-series data of variables related to healthcare for each user. Since the relationships between variables are newly calculated by performing the modified analysis process based on the series data, time series data of variables related to the user's health care can be analyzed with high accuracy. As a result, highly accurate analysis results that match or approximate the results of AI (artificial intelligence) machine learning can be quickly calculated on the local user information terminal device, thereby improving the user's health care system. It will be possible to contribute to its spread.

Further, the user-side analysis processing section and the server-side analysis processing section may perform a plurality of analysis processes. According to this, in a user information terminal device, relationships between variables are calculated by performing multiple analysis processes based on time-series data of variables related to the user's health care. The accuracy of time-series data analysis can be improved.

Furthermore, the user-side analysis processing section and the server-side analysis processing section may set weighting values for each analysis process. According to this, by combining each analysis process and the weighting value, it is possible to easily and reliably design a predetermined analysis process.

Further, in the server device, the analysis processing changing unit may be configured to adjust the first relationship between the variables related to each user's healthcare calculated by the first server-side relationship calculation unit to the second server-side relationship calculation unit. Changing the weighting value of each analysis process in the first server-side relationship calculation unit so as to match or approximate the second relationship between the variables related to each user's healthcare calculated by the relationship calculation unit. The server-side transmission unit transmits change information regarding the weighting value of each analysis process changed by the analysis process change unit to the user information terminal device via the network, and in the user information terminal device, the The user-side receiving unit receives change information regarding the numerical value of weighting of each analysis process transmitted from the server device via the network, and the user-side analysis processing unit receives each change information received by the user-side receiving unit. After changing the weighting value of each analysis process based on change information regarding the weighting value of the analysis process, and after changing it based on the time series data of variables related to the user's healthcare acquired by the data acquisition unit. The user-side relationship calculation unit calculates the relationship between variables related to the user's health care based on the results of each analysis process after changes performed by the user-side analysis processing unit. It may be newly calculated. According to this, the results of the analysis process can be corrected simply by changing the weighting values for each analysis process, making it possible to easily and reliably improve the analysis accuracy of time-series data of variables related to the user's healthcare. .

Further, the user-side relationship calculation unit, the first server-side relationship calculation unit, and the second server-side relationship calculation unit calculate the relationship between variables related to the user's health care by calculating the relationship between one variable and another variable. It is also possible to calculate a ranking of the relationships between the two. According to this, the ranking of the relationship between one variable and another variable is calculated as the relationship between variables related to the user's health care, so the user can easily understand the relationship between variables related to his or her health care. can do.

The present invention also provides a user information terminal device and a server device connected via a network, which are used in the healthcare data analysis system described above, which analyzes time-series data of variables related to the user's healthcare. The information terminal device includes a data acquisition unit that acquires time series data of variables related to the user's health care, and a predetermined analysis based on the time series data of the variables related to the user's health care acquired by the data acquisition unit. a user-side analysis processing unit that performs processing; and a user-side relationship calculation unit that calculates a relationship between variables related to the user's health care based on the results of a predetermined analysis process performed by the user-side analysis processing unit. a user-side transmitting unit that transmits time-series data of variables related to the user's health care acquired by the data acquiring unit to the server device via a network; and a user-side receiving unit that receives change information regarding a predetermined analysis process.

The present invention also provides a server for use in the healthcare data analysis system described above, in which a plurality of user information terminal devices and a server device are connected via a network, and which analyzes time-series data of variables related to the user's healthcare. The apparatus includes: a server-side receiving section that receives time-series data of variables related to healthcare of each user transmitted from each user information terminal via a network; A server-side analysis processing unit that performs a predetermined analysis process based on time-series data of variables related to healthcare; and a server-side analysis processing unit that performs predetermined analysis processing based on time-series data of variables related to healthcare; and based on the results of the analysis processing performed by the server-side analysis processing unit, a first server-side relationship calculation unit that calculates the relationship between the two; a learning unit that performs machine learning based on time-series data of variables related to healthcare of each user received by the server-side reception unit; A second server-side relationship calculation unit that calculates relationships between variables related to healthcare for the entire user based on the results of machine learning performed by the unit, and the first server-side relationship calculation unit. the relationship between variables related to health care for all users that has been calculated matches or approximates the relationship between variables related to health care for all users calculated by the second server-side relationship calculation unit; an analysis process change unit that changes a predetermined analysis process in the server-side relationship calculation unit of No. 1; and transmits change information regarding the predetermined analysis process changed by the analysis process change unit to the user information terminal device via a network. and a server-side transmission unit.

The present invention also provides a computer for use in the healthcare data analysis system described above, in which a plurality of user information terminal devices and a server device are connected via a network, and which analyzes time-series data of variables related to the user's healthcare. a data acquisition unit that acquires time-series data of variables related to the user's health care; a user-side analysis processing unit that performs predetermined analysis processing based on the user-side analysis processing unit; and a user-side relationship that calculates a relationship between variables related to the user's health care based on the results of the predetermined analysis processing performed by the user-side analysis processing unit. a user-side transmission unit that transmits time-series data of variables related to health care of the user acquired by the data acquisition unit to the server device via a network; It is characterized in that it functions as a user-side receiving unit that receives change information regarding a predetermined analysis process.

The present invention also provides a computer for use in the healthcare data analysis system described above, in which a plurality of user information terminal devices and a server device are connected via a network, and which analyzes time-series data of variables related to the user's healthcare. The program includes a server-side receiving unit that receives time-series data of variables related to healthcare of each user transmitted from each user information terminal via a network, and is received by the server-side receiving unit. A server-side analysis processing unit that performs predetermined analysis processing based on time-series data of variables related to healthcare for each user; a first server-side relationship calculation unit that calculates a relationship between variables related to the above; a learning unit that performs machine learning based on time-series data of variables related to each user's healthcare received by the server-side reception unit; A second server-side relationship calculation unit that calculates relationships between variables related to healthcare for the entire user based on the results of machine learning performed by the learning unit, and calculated by the first server-side relationship calculation unit. the relationship between variables related to health care for all users that has been calculated matches or approximates the relationship between variables related to health care for all users calculated by the second server-side relationship calculation unit; an analysis process change unit that changes a predetermined analysis process in the server-side relationship calculation unit of No. 1, and transmits change information regarding the predetermined analysis process changed by the analysis process change unit to the user information terminal device via a network; It is characterized by functioning as a server-side transmitter.

According to the present invention, in a user information terminal device owned or used by a user, relationships between variables are calculated by performing a predetermined analysis process based on time-series data of variables related to the user's health care. It is possible to quickly analyze time-series data of variables related to the user's health care. In addition, the server device collects time-series data of variables related to each user's healthcare, and after changing a predetermined analysis process using machine learning, the user information terminal device collects time-series data of variables related to healthcare for each user. Since the relationships between variables are newly calculated by performing the modified analysis process based on the series data, time series data of variables related to the user's health care can be analyzed with high accuracy. As a result, highly accurate analysis results that match or approximate the results of AI (artificial intelligence) machine learning can be quickly calculated on the local user information terminal device, thereby improving the user's health care system. It will be possible to contribute to its spread.

1 is a diagram showing the overall configuration of a healthcare data analysis system according to the present embodiment. FIG. 2 is a block diagram showing the configuration of a user-side information terminal and a server device in FIG. 1. FIG. 2 is a diagram showing an overview of the healthcare data analysis system of FIG. 1. FIG. 2 is a flowchart showing the operation of the healthcare data analysis system of FIG. 1. FIG. FIG. 3 is a diagram showing an example of a screen displayed on a user-side information terminal device. FIG. 7 is a diagram illustrating another example of a screen displayed on a user-side information terminal device.

Next, an embodiment of a healthcare data analysis system (hereinafter referred to as the present system) according to the present invention will be described with reference to FIGS. 1 to 5.

As shown in FIG. 1, this system includes a plurality of user information terminal devices 1 such as smartphones, tablet terminals, or wearable terminals, and a server device 2 that collects data from each user information terminal device 1. A terminal device 1 and a server device 2 are communicably connected via a network such as the Internet.

[Configuration of user information terminal device 1]
The user information terminal device 1 is a device such as a smartphone, a tablet terminal, or a wearable device that can be connected to a network, and includes time-series data of variables related to the health care of the user who is the owner or user (for example, the life cycle data shown in FIG. 3). By performing an analysis process on the log data (100), relationships between variables related to the user's health care are calculated.

Specifically, as shown in FIG. 2, the user information terminal device 1 includes a data acquisition unit 11 that acquires time series data of variables related to the user's health care, and a data acquisition unit 11 that acquires time series data of variables related to the user's health care. A data storage unit 12 that stores data; a user-side data processing unit 13 that processes time-series data of variables related to the user's health care; a user-side analysis processing unit 14 that performs analysis processing; a user-side relationship calculation unit 15 that calculates relationships between variables related to the user's health care; and an output unit that outputs relationships between variables related to the user's health care. 16, a user-side transmitter 17 that transmits time-series data of variables related to the user's health care to the server device 2 via the network, and changes regarding predetermined analysis processing transmitted from the server device 2 via the network. and a user-side receiving section 18 that receives information.

The data acquisition unit 11 acquires time-series data of variables related to healthcare such as the user's (individual) body, health, activity, or environment from outside or inside the terminal, and is measured by an IoT device such as a wearable device. time-series data of variables measured or input from an external web service, and time-series data of variables measured or input using an application on an information terminal device.

Time-series data on variables related to healthcare for these users include climate time-series (temperature, atmospheric pressure, weather, moon phase, etc.), environmental condition time-series (location, room temperature, humidity, music, etc.), location categorical (movement, office , home, hobbies, etc.), habit categorical (meals, exercise, hygiene, drinking, etc.), habit time series (step count, calories, desk work, sleep, etc.), vital data time series (heart rate, blood pressure, body temperature, weight, etc.) , time-series data of each variable belonging to categorical physical condition (good, fair, tired, bad, etc.). Among these, time series data of variables related to location categorical, habit categorical, habit time series, vital data time series, and physical condition categorical are often obtained from IoT devices. Further, time-series data regarding variables such as climate time-series and environmental condition time-series are often obtained from external web services. Furthermore, time series data of variables related to vital data time series may be acquired from the user information terminal device 1.

In addition, the data acquisition unit 11 acquires variable time series data using a data automatic input plug-in in applications such as a health kit, a calendar, external weather data, a sleep application, and a life log application installed in the user information terminal device 1. Automatically or manually enter time-series data on variables related to physical condition, exercise, hygiene, diet, location, custom settings, notes, etc. For example, as an example of acquiring time series data of variables by manual input, you can set the unit of 1 routine yourself and enter it in units of 0.5 routine, 1.5 routine, etc. You can input the time of your habit, or input your physical condition for the day (or the symptoms if you feel unwell).

In addition, in this embodiment, the following is mentioned as specific time-series data of variables related to the user's healthcare acquired by the data acquisition unit 11. However, the acquisition of time series data of each of these variables is just an example, and time series data of various other types of variables may be acquired.

・asleep: Sleep time ・temperature_min: Lowest temperature of the day ・Efficiency: Sleep efficiency (sleep time/time between going to bed and waking up)
・wakingBPM_diff: Daily change (difference) in heart rate (Beats per minute) when waking up
・quality: Sleep quality ・dayBPM: Daytime heart rate (Beats per minute)
・wakingBPM: Heart rate when waking up (Beats per minute)
・pressure_noon_diff: Daily change in atmospheric pressure (difference)
・hrv: Change in heartbeat interval ・deep: Deep sleep time ・asleep_diff: Daily change in sleep time (difference)

The user-side data processing section 13 performs a predetermined process on time-series data of variables related to the user's health care stored in the data storage section 12 for a predetermined analysis process by the user-side analysis processing section 14, which will be described later. This processing process is performed. For example, the user-side data processing unit 13 performs a process of converting json data into csv data regarding hourly or daily time series data related to the user's healthcare stored in the data storage unit 12. , reads in a format suitable for data processing (eg, DataFrame, array, list, etc.). In addition. If the date format is different, perform the processing necessary to read the time series data at this point. In addition, the user-side data processing unit 13 analyzes time-series data related to personal healthcare stored in the data storage unit 12 that may affect variables related to healthcare that are the main targets of analysis processing. Filter only time series data for variables. In addition, when the user-side analysis processing section 14 (to be described later) performs time correlation and DTW (Dynamic Time Warping) analysis processing, the user-side data processing section 13 performs linear interpolation, 0 interpolation, and 0 interpolation. &Performs missing interpolation processing such as EMA.

The user-side analysis processing section 14 performs a predetermined analysis process based on the time-series data of variables related to the user's health care processed by the user-side data processing section 13 . As this analysis process, for example, an analysis process regarding the similarity and correlation of time series data of each variable is performed using a known algorithm such as time correlation or DTW (Dynamic Time Warping).

Further, the user-side analysis processing section 14 is configured to perform a plurality of analysis processes, and a weighting value is set for each analysis process. For example, the user-side analysis processing unit 14 is configured to perform three analysis processes A, B, and C, as shown in (101) in FIG. 3, and the initial setting value of the weighting of analysis process A is 0. 33. The initial setting value of the weighting of analysis processing B is set to 0.33, and the initial setting value of the weighting of analysis processing C is set to 0.33. At the initial stage, the weighting of analysis processing A, B, and C is The values are manually set evenly.

In addition, the user-side analysis processing unit 14 changes a predetermined analysis process based on change information regarding the predetermined analysis process received by a user-side reception unit 18 (described later), and then changes a variable related to the user's health care. A predetermined analysis process after the change is performed based on the time series data. For example, as shown in (103) in FIG. Weighting values for analysis processes A, B, and C (weighting values for analysis process A: 0.33 → 0.4, weighting values for analysis process B: 0.33 → 0.35, weighting values for analysis process C) (Number: 0.33 → 0.25) periodically or sequentially.

The user-side relationship calculation unit 15 calculates relationships between variables related to the user's health care based on the results of the analysis process performed by the user-side analysis processing unit 14. For example, as shown in (101) in FIG. 3, the user-side relationship calculation unit 15 calculates one of Ranking of relationships between other variables (explanatory variables: α, β, γ, θ, ...) with respect to variables (objective variables) (1 = data α, 2 = data γ, 3 = data θ, 4 = data β, ... ) is calculated.

In addition, when the user-side analysis processing section 14 changes the analysis processing, the user-side relationship calculation section 15 calculates the result based on the result of the changed predetermined analysis processing performed by the user-side analysis processing section 14. Then, the relationship between variables related to the user's health care is newly calculated. For example, as shown in (104) in FIG. Based on the results of each analysis process A, B, and C in which the weighting values performed by the processing unit 14 have been changed, the relationship between variables related to the user's health care (1 = data α, 2 = data β, 3=data γ, 4=data θ,...) are newly calculated.

The output unit 16 outputs the relationship between variables related to the user's health care calculated by the user-side relationship calculation unit 15. For example, as shown in FIG. (headache), weather, temperature, event (bath)...) is displayed on the screen.

The user-side transmitting unit 17 transmits time-series data of variables related to the user's health care stored in the data storage unit 12 to the server device 2 via the network. Since the time-series data of this variable is related to the personal information of the user, it is transmitted without identifying the individual.

The user-side receiving unit 18 receives change information regarding a predetermined analysis process transmitted from the server device 2 via the network. In this embodiment, as will be described later, change information regarding the weighting values of each analysis process A, B, and C is received from the server device 2 via the network.

[Configuration of server device 2]
The server device 2 is a server computer connected to a network, and collects time-series data (for example, big data 200 in FIG. 3) of variables related to each user's health care from each user information terminal device 1. By performing machine learning using artificial intelligence, relationships between variables related to healthcare are calculated. Note that the server device 2 does not calculate the relationship between variables related to health care for each user, but calculates the relationship between variables related to health care for all users.

Specifically, the server device 2 includes a server-side receiving unit 21 that receives time-series data of variables related to each user's health care transmitted from each user information terminal device 1 via the network, and A first server-side data processing section 22 that processes time-series data of variables related to healthcare; and a server-side analysis processing section that performs predetermined analysis processing based on the time-series data of variables related to healthcare of each user. 23, and a first server-side relationship calculation unit 24 that calculates a first relationship between variables related to each user's health care based on a predetermined analysis process.

The server device 2 also includes a second server-side data processing unit 25 that processes time-series data of variables related to each user's health care received by the server-side receiving unit 21, and A learning unit 26 that performs machine learning based on time-series data of variables related to care, and a second server-side relationship calculation that calculates a second relationship between variables related to each user's health care based on the machine learning results. 27.

Furthermore, the server device 2 includes an analysis processing changing unit 28 that changes the predetermined analysis processing in the first server-side relationship calculation unit 24, and a change information regarding the predetermined analysis processing to the user information terminal device 1 via the network. and a server-side transmitter 29 for transmitting data.

The server-side receiving unit 21 receives time-series data of variables related to each user's health care transmitted from each user information terminal via the network. These time-series data of variables related to each user's health care are related to each user's personal information, and therefore are received without identifying the individual.

The first server-side data processing unit 22 processes the time-series data of variables related to healthcare of each user received by the server-side reception unit 21 for predetermined analysis processing by the server-side analysis processing unit 23, which will be described later. Predetermined processing is performed on the material. This predetermined processing is generally the same as the processing performed by the user-side data processing unit 13 of the user information terminal device 1, and therefore the description thereof will be omitted.

The server-side analysis processing unit 23 performs a predetermined analysis process based on the time-series data of variables related to healthcare of each user that has been processed by the first server-side data processing unit 22 . This analysis processing is the same analysis processing as the user-side analysis processing section 14 in the user information terminal device 1, and uses known algorithms such as time correlation and DTW (Dynamic Time Warping) to calculate each variable. Analyzes the similarity and correlation of time series data.

Further, the server-side analysis processing unit 23 is configured to perform a plurality of analysis processes, similar to the user-side analysis processing unit 14 in the user information terminal device 1, and a weighting value is set for each analysis process. . For example, the server-side analysis processing unit 23 is configured to perform three analysis processes A, B, and C, as shown in (201) in FIG. 3, and the initial setting value of the weighting of analysis process A is 0.33. , the initial setting value of the weighting of the analysis processing B is set to 0.33, and the initial setting value of the weighting of the analysis processing C is set to 0.33.

Furthermore, when a predetermined analysis process is changed by an analysis process change unit 28 (described later), the server-side analysis processing unit 23 thereafter changes the predetermined change based on time-series data of variables related to each user's health care. Perform analysis processing. For example, as shown in (201) in FIG. 3, the server-side analysis processing unit 23 uses the analysis processing change unit 28 to set the weighting values of each analysis process A, B, and C (weighting value of analysis process A: 0. 33 → 0.4, the weighting value of analysis process B: 0.33 → 0.35, the weighting value of analysis process C: 0.33 → 0.25), so after that the health of each user Each analysis process A, B, and C is performed in which the weighting values are changed based on time-series data of variables related to care.

The first server-side relationship calculation unit 24 calculates a first relationship between variables related to each user's health care based on the result of the analysis process performed by the server-side analysis processing unit 23. For example, as shown in (202) in FIG. , the first relationship (ranking: 1 = data α, 2 = data γ, 3 = data θ, 4 = data β,…) is calculated.

In addition, when the analysis processing is changed in the server-side analysis processing unit 23, the first server-side relationship calculation unit 24 calculates a change in the predetermined analysis processing performed by the server-side analysis processing unit 23. Based on the results, a first relationship between variables related to each user's health care is newly calculated. For example, in the present embodiment, when the weighting values of each analysis process A, B, and C are changed in the server-side analysis process unit 23, the first server-side relationship calculation unit 24 calculates that Based on the results of each analysis process in which the weighting values by the side analysis processing unit 23 have been changed, a first relationship (ranking) between variables related to healthcare for each user as a whole is newly calculated.

The second server-side data processing unit 25 performs prior organization and processing as big data that can be subjected to machine learning by a learning unit 26, which will be described later. For example, the second server-side data processing unit 25 may perform lag processing on the time-series data of a plurality of variables received from each user-side information terminal device 1 according to the characteristics (for example, sleep data may be -1 (e.g., 30 days), perform average processing over a period of N days (e.g., 2-day average, 3-day average, etc.), limit data to user-side information terminals that have data for a certain period of time (e.g., 30 days), Predetermined variables are selected by a predetermined algorithm (eg, hierarchical clustering, correlation coefficient, VIF).

The learning unit 26 uses AI (Artificial Intelligence) to develop an optimal machine based on time-series data of variables related to healthcare of each user processed by the second server-side data processing unit 25. Learn. For example, as shown in (203) in FIG. 3, the learning unit 26 selects machine learning with high prediction accuracy using time series data (big data) of each user, and determines which explanatory variables contributed to the objective variable. is calculated using multiple XAI techniques.

Note that examples of machine learning in the learning unit 26 include Feature Importance (gain and split), SHAP, and Permutation Importance. "Feature Importance" is an index that evaluates the classification contribution rate of the target of each feature (corresponding to a variable in the present invention), and is a standard of how much Gini impurity can be reduced by dividing by a certain feature. be. In addition, "SHAP (SHApley Additive exPlanations)" is an open source library that applies the Shapley Value of cooperative game theory to machine learning. In order to solve the problem that the number of combinations increases and the amount of calculation becomes enormous when , the calculation method is devised so that Shapley values can be handled by machine learning in a realistic calculation time. Furthermore, "Permutation Importance" is one of the methods to measure the usefulness of the features of a machine learning model, and a commonly used method is Feature Importance (for LightGBM), which is a method for measuring the usefulness of features in a machine learning model. Quantify and express how well the segmentation is going for each feature (variable). In addition, since the learning unit 26 handles the granularity of variables such as physical condition and sleep time in the user-side analysis processing unit 14 and the server-side analysis processing unit 23, the learning unit 26 calculates the sum and average as a process to match variables (features) to the same granularity. May be used.

The second server-side relationship calculation unit 27 calculates a second relationship between variables related to each user's health care based on the results of machine learning performed by the learning unit 26. For example, as shown in (204) in FIG. 3, the second server-side relationship calculation unit 27 calculates the time series of variables using AI technology such as Feature Importance (gain and split), SHAP, and Permutation Importance by the learning unit 26. Based on the results of data machine learning, the second relationship (ranking: 1 = data α, 2 = data β, 3 = data γ, 4 = data θ, ...) are calculated.

The analysis processing change unit 28 changes the first relationship between the variables related to each user's healthcare calculated by the first server-side relationship calculation unit 24 to the second relationship calculation unit 27 . The predetermined analysis process in the first server-side relationship calculation unit 24 is changed so that the second relationship of variables related to healthcare of each user is matched or approximated. For example, as shown in (202) and (204) in FIG. The relationship (ranking: 1 = data α, 2 = data γ, 3 = data θ, 4 = data β, ...) is the relationship between variables related to each user's healthcare calculated by the second server-side relationship calculation unit 27. As shown in (201) in FIG. 3, in order to match or approximate the second relationship (ranking: 1 = data α, 2 = data β, 3 = data γ, 4 = data θ, ...), The weighting values for each analysis process in the first server-side relationship calculation unit 24 (for example, the weighting value for analysis process A: 0.33 → 0.4, the weighting value for analysis process B: 0.33 → 0 .35, the weighting value of analysis processing C: 0.33 → 0.25) is changed.

The server-side transmitter 29 transmits change information regarding the predetermined analysis process changed by the analysis process changer 28 (for example, the weighting value of analysis process A: 0.33 → 0.4, the weighting value of analysis process B). :0.33→0.35, weighting value of analysis processing C:0.33→0.25) is transmitted to the user information terminal device 1 via the network.

[Operation of this system]
Next, the operation of this system will be explained with reference to the flowchart of FIG.

First, in the user information terminal device 1, the data acquisition unit 11 acquires time-series data of variables related to healthcare such as the user's (individual) body, health, activity, or environment (S11).

Then, the data storage unit 12 stores time-series data of variables related to the user's health care acquired by the data acquisition unit 11 (S12).

Then, the user-side data processing unit 13 processes the time-series data of variables related to the user's healthcare stored in the data storage unit 12 for predetermined analysis processing by the user-side analysis processing unit 14, which will be described later. Predetermined processing is performed (S13).

Then, the user-side analysis processing unit 14 performs a predetermined analysis process based on the time-series data of variables related to the user's healthcare processed by the user-side data processing unit 13 (S14).

Then, the user-side relationship calculation unit 15 calculates the relationship (ranking) between variables related to the user's health care based on the result of the analysis process performed by the user-side analysis processing unit 14 (S15).

Then, the output unit 16 outputs the relationship (ranking) of variables related to the user's health care calculated by the user-side relationship calculation unit 15 (S16).

Further, the user-side transmitting unit 17 transmits time-series data of variables related to the user's healthcare stored in the data storage unit 12 to the server device 2 via the network (S17).

Next, in the server device 2, the server-side receiving unit 21 receives time-series data of variables related to each user's health care transmitted from the user information terminal device 1 via the network (S21).

The first server-side data processing unit 22 then processes time-series data of variables related to healthcare of each user received by the server-side reception unit 21 for predetermined analysis processing by the server-side analysis processing unit 23, which will be described later. A predetermined processing process is performed on the image (S22).

Then, the server-side analysis processing unit 23 performs a predetermined analysis process based on the time-series data of variables related to each user's healthcare that has been processed by the first server-side data processing unit 22 (S23). .

Then, the first server-side relationship calculation unit 24 calculates a first relationship (ranking) between variables related to each user's healthcare based on the results of the analysis process performed by the server-side analysis processing unit 23. Calculate (S24).

On the other hand, also in the server device 2, the second server-side data processing unit 25 performs prior organization and processing as big data that can be subjected to machine learning by the learning unit 26 (described later) (S25).

Then, the learning unit 26 uses AI (artificial intelligence) to perform optimal machine learning based on the time series data of variables related to healthcare of each user that has been processed by the second server-side data processing unit 25. Execute (S26).

Then, the second server-side relationship calculation unit 27 calculates a second relationship (ranking) between variables related to each user's healthcare based on the results of machine learning performed by the learning unit 26 ( S27).

The analysis processing change unit 28 then determines that the first relationship (ranking) between the variables related to each user's healthcare calculated by the first server-side relationship calculation unit 24 is based on the second server-side relationship. Each analysis process A, B, The weighting value of C is changed (S28).

Then, the server-side transmitting unit 29 transmits change information regarding the predetermined analysis process changed by the analysis process change unit 28 (for example, the weighting value of analysis process A is 0.33 → 0.4, the weighting value of analysis process B is The numerical value is 0.33→0.35, and the numerical value of the weighting of analysis process C is 0.33→0.25) is transmitted to the user information terminal device 1 via the network (S29).

Next, in the user information terminal device 1, the user-side receiving unit 18 receives change information regarding the weighting values of each of the analysis processes A, B, and C transmitted via the network from the server device 2 (S18). .

Then, the user-side analysis processing unit 14 adjusts the weighting of each analysis process A, B, and C based on change information regarding the weighting value of each analysis process A, B, and C received by the user-side reception unit 18 (described later). The numerical value is changed (S19).

After that, the user information terminal device 1 performs the processes of S11 to S16 to perform each of the changed analysis processes A, B, and C based on the time-series data of variables related to the user's health care. The relationship (ranking) between variables related to healthcare is newly calculated and output.

In this embodiment, the user-side analysis processing unit 14 and the server-side analysis processing unit 23 set weighting values for each analysis process, but they may not set weighting values. Further, although the user-side analysis processing unit 14 and the server-side analysis processing unit 23 are assumed to perform a plurality of analysis processes, they may also perform one analysis process. Furthermore, the user-side analysis processing section 14 and the server-side analysis processing section 23 are not limited to the above-mentioned time correlation and DTW algorithms, and may perform analysis processing using other algorithms; From the viewpoint of speed, large-scale analysis processing such as machine learning using AI is not included.

In addition, the user-side analysis processing unit 14 and the server-side analysis processing unit 23 use the “relationship estimation system” of Japanese Patent Application Laid-Open No. 2022-013409, which was previously proposed by the applicant, as one of the predetermined analysis processes. May be used. To explain this relationship estimation system specifically, in the relationship estimation system, an information terminal device collects time-series data of multiple types of attributes (variables) from an IoT device external to the terminal, an external web service, and an application on the terminal. , calculate the similarity between the time series data of each attribute (variable), use this similarity as the weight of the link between each node, and generate an undirected relationship graph in which each node is connected by a link. Routes with large link weights are calculated for each node, and based on the plurality of routes, nodes with large overall link weights are extracted. Then, the correlation between the time series data of each attribute (variable) is calculated, and the causal relationship between each node of the relationship graph according to each attribute (variable) is calculated, and based on the causal relationship between each node. Then, a directed relationship graph is updated, and relationships between attributes (variables) corresponding to the extracted nodes are estimated.

Further, as shown in FIG. 5, the output unit 16 of the user information terminal device 1 is configured to output a plurality of other explanatory variables for the objective variable in the form of a ranking, but the output format is not limited. Instead, for example, as shown in FIG. 6, one explanatory variable for the target variable may be output as detailed information such as a graph.

Although the embodiments of the present invention have been described above with reference to the drawings, the present invention is not limited to the illustrated embodiments. Various modifications and variations can be made to the illustrated embodiments within the same scope as the present invention or within an equivalent scope.

1... User information terminal device 11... Data acquisition section 12... Data storage section 13... User side data processing section 14... User side analysis processing section 15... User side relationship calculation section 16... Output section 17... User side transmission section 18 ...User-side receiving unit 2...Server device 21...Server-side receiving unit 22...First server-side data processing unit 23...Server-side analysis processing unit 24...First server-side relationship calculation unit 25...Second server Side data processing unit 26...Learning unit 27...Second server side relationship calculation unit 28...Analysis processing change unit 29...Server side transmission unit

Claims (9)

A healthcare data analysis system in which a plurality of user information terminal devices and a server device are connected via a network to analyze time-series data of variables related to the user's healthcare, the system comprising:
The user information terminal device includes:
a data acquisition unit that acquires time series data of variables related to the user's healthcare;
a user-side analysis processing unit that performs predetermined analysis processing based on time-series data of variables related to the user's healthcare acquired by the data acquisition unit;
a user-side relationship calculation unit that calculates a relationship between variables related to the user's health care based on the results of a predetermined analysis process performed by the user-side analysis processing unit;
a user-side transmission unit that transmits time-series data of variables related to the user's health care acquired by the data acquisition unit to the server device via a network;
a user-side receiving unit that receives change information regarding a predetermined analysis process transmitted from the server device via a network;
The server device includes:
a server-side receiving unit that receives time-series data of variables related to each user's healthcare transmitted from each user information terminal device via the network;
a server-side analysis processing unit that performs predetermined analysis processing based on time-series data of variables related to healthcare of each user received by the server-side reception unit;
a first server-side relationship calculation unit that calculates a first relationship between variables related to healthcare for the entire user based on the result of the analysis process performed by the server-side analysis process unit;
a learning unit that performs machine learning based on time-series data of variables related to healthcare of each user received by the server-side receiving unit;
a second server-side relationship calculation unit that calculates a second relationship between variables related to healthcare for the entire user based on the results of machine learning performed by the learning unit;
The first relationship between variables related to health care for all users calculated by the first server-side relationship calculation unit is related to health care for all users calculated by the second server-side relationship calculation unit. an analysis process change unit that changes a predetermined analysis process in the first server-side relationship calculation unit so as to match or approximate a second relationship between variables;
a server-side transmitting unit that transmits change information regarding a predetermined analysis process changed by the analysis process change unit to the user information terminal device via a network;
In the user information terminal device,
The user-side analysis processing unit changes the predetermined analysis process based on the change information regarding the predetermined analysis process received by the user-side reception unit, and then changes the user's health acquired by the data acquisition unit. Perform predetermined analysis processing after changes based on time-series data of variables related to care,
The user-side relationship calculation unit newly calculates a relationship between variables related to the user's health care based on the result of a predetermined analysis process after the change performed by the user-side analysis processing unit. Characteristic data analysis system for healthcare. The healthcare data analysis system according to claim 1, wherein the user-side analysis processing unit and the server-side analysis processing unit perform a plurality of analysis processes. 3. The healthcare data analysis system according to claim 2, wherein the user-side analysis processing section and the server-side analysis processing section set weighting values for each analysis process. In the server device,
The analysis processing change unit is configured to calculate a first relationship between variables related to each user's health care calculated by the first server-side relationship calculation unit, and a first relationship calculated by the second server-side relationship calculation unit. changing the weighting value of each analysis process in the first server-side relationship calculation unit so as to match or approximate a second relationship between variables related to healthcare of each user;
The server-side transmission unit transmits change information regarding the weighting value of each analysis process changed by the analysis process change unit to the user information terminal device via a network,
In the user information terminal device,
The user-side receiving unit receives change information regarding the numerical value of weighting of each analysis process transmitted from the server device via the network,
The user-side analysis processing unit changes the weighting value of each analysis process based on change information regarding the weighting value of each analysis process received by the user-side reception unit, and then changes the weighting value of each analysis process, and then changes the weighting value of each analysis process, and then changes the weighting value of each analysis process based on the change information regarding the weighting value of each analysis process received by the user-side reception unit Perform each analysis process after changes based on the time-series data of variables related to the user's healthcare,
3. The user-side relationship calculation unit newly calculates the relationship between variables related to the user's health care based on the results of each changed analysis process performed by the user-side analysis processing unit. A data analysis system for healthcare described in . The user-side relationship calculation unit, the first server-side relationship calculation unit, and the second server-side relationship calculation unit calculate the relationship between one variable and another variable as a relationship between variables related to the user's health care. The healthcare data analysis system according to any one of claims 1 to 4, which calculates a gender ranking. The user information terminal device used in the healthcare data analysis system according to claim 1, wherein a plurality of user information terminal devices and a server device are connected via a network, and analyze time-series data of variables related to the user's healthcare. There it is,
a data acquisition unit that acquires time series data of variables related to the user's healthcare;
a user-side analysis processing unit that performs predetermined analysis processing based on time-series data of variables related to the user's healthcare acquired by the data acquisition unit;
a user-side relationship calculation unit that calculates a relationship between variables related to the user's health care based on the results of a predetermined analysis process performed by the user-side analysis processing unit;
a user-side transmission unit that transmits time-series data of variables related to the user's health care acquired by the data acquisition unit to the server device via a network;
A user information terminal device comprising: a user-side receiving unit that receives change information regarding a predetermined analysis process transmitted from the server device via a network. 2. The server device used in the healthcare data analysis system according to claim 1, wherein the plurality of user information terminal devices and the server device are connected via a network, and analyze time-series data of variables related to the user's healthcare. ,
a server-side receiving unit that receives time-series data of variables related to each user's healthcare transmitted from each user information terminal via the network;
a server-side analysis processing unit that performs predetermined analysis processing based on time-series data of variables related to healthcare of each user received by the server-side reception unit;
a first server-side relationship calculation unit that calculates relationships between variables related to healthcare for the entire user based on the results of the analysis processing performed by the server-side analysis processing unit;
a learning unit that performs machine learning based on time-series data of variables related to healthcare of each user received by the server-side receiving unit;
a second server-side relationship calculation unit that calculates relationships between variables related to healthcare for all users based on the results of machine learning performed by the learning unit;
The relationship between the variables related to healthcare for all users calculated by the first server-side relationship calculation unit is the same as the relationship between variables related to healthcare for all users calculated by the second server-side relationship calculation unit. an analysis process change unit that changes a predetermined analysis process in the first server-side relationship calculation unit so as to match or approximate the relationship;
A server device comprising: a server-side transmitter that transmits change information regarding a predetermined analysis process changed by the analysis process changer to the user information terminal device via a network. 2. A computer program for use in a healthcare data analysis system according to claim 1, wherein a plurality of user information terminal devices and a server device are connected via a network and analyze time-series data of variables related to the user's healthcare. ,
The user information terminal device,
a data acquisition unit that acquires time series data of variables related to the user's healthcare;
a user-side analysis processing unit that performs predetermined analysis processing based on time-series data of variables related to the user's healthcare acquired by the data acquisition unit;
a user-side relationship calculation unit that calculates a relationship between variables related to the user's health care based on the results of a predetermined analysis process performed by the user-side analysis processing unit;
a user-side transmission unit that transmits time-series data of variables related to the user's health care acquired by the data acquisition unit to the server device via a network;
A computer program that functions as a user-side receiving unit that receives change information regarding a predetermined analysis process transmitted from the server device via a network. 2. A computer program for use in a healthcare data analysis system according to claim 1, wherein a plurality of user information terminal devices and a server device are connected via a network and analyze time-series data of variables related to the user's healthcare. ,
The server device,
a server-side receiving unit that receives time-series data of variables related to each user's healthcare transmitted from each user information terminal via the network;
a server-side analysis processing unit that performs predetermined analysis processing based on time-series data of variables related to healthcare of each user received by the server-side reception unit;
a first server-side relationship calculation unit that calculates relationships between variables related to healthcare for the entire user based on the results of the analysis process performed by the server-side analysis process unit;
a learning unit that performs machine learning based on time-series data of variables related to healthcare of each user received by the server-side receiving unit;
a second server-side relationship calculation unit that calculates relationships between variables related to healthcare for all users based on the results of machine learning performed by the learning unit;
The relationship between the variables related to healthcare for all users calculated by the first server-side relationship calculation unit is the same as the relationship between variables related to healthcare for all users calculated by the second server-side relationship calculation unit. an analysis process changing unit that changes a predetermined analysis process in the first server-side relationship calculation unit so as to match or approximate the relationship;
A computer program that functions as a server-side transmitter that transmits change information regarding a predetermined analysis process changed by the analysis process changer to the user information terminal device via a network.