Computer program, information processing apparatus, and method

Abstract

Provided are a computer program, an information processing device, and a method for inferring a state of health of a subject user based on a history of behaviors performed by the subject user and having at least partially improved performance. A computer program according to one embodiment is capable of causing at least one processor to perform the following functions by being executed by the at least one processor: acquiring object behavior data for identifying a history of behaviors performed by an object user, inputting the subject behavior data into an inference model generated by performing supervised learning to output subject sleep state data for identifying a sleep state of the subject user from the inference model, the object sleep state data comprises object property data used for identifying the sleep property of the object user, object rhythm data used for identifying the sleep rhythm of the object user, and object time data used for identifying the sleep time of the object user; and at least one of object type data for identifying a sleep type to which the object user belongs.

Description

Technical Field

[0001] The technology disclosed in the present application relates to a computer program, an information processing device, and a method for inputting data for identifying a historical record of behaviors performed by a user set as an inference object (target user) into a learning model and outputting data for identifying the health status of the target user from the above-mentioned learning model. Background Art

[0002] Japanese Patent No. 6916367 (Patent Document 1) discloses a method of constructing an inference model for inferring the health status of a target user based on data describing the purchase history of the target user. The entirety of this patent document is incorporated into this specification by reference.

[0003] Prior art literature

[0004] Patent Literature

[0005] Patent Document 1: Japanese Patent No. 6916367 Summary of the invention

[0006] Problem that the invention aims to solve

[0007] Recently, as a method of inferring the health status of a target user based on a history of behaviors performed by the target user, there is a demand for providing a method having further improved performance.

[0008] Thus, the technology disclosed in the present application provides a computer program, an information processing device, and a method for inferring the health status of a target user based on a history of behaviors performed by the target user and having at least partially improved performance.

[0009] Methods used to solve problems

[0010] A computer program involved in one embodiment is capable of "being executed by at least one processor to cause the at least one processor to perform the following functions: obtaining object behavior data for identifying a historical record of behaviors performed by an object user, and outputting object sleep state data for identifying a sleep state of the object user from the inference model by inputting the object behavior data into an inference model generated by performing supervised learning, the object sleep state data including at least one of object property data for identifying a sleep property of the object user, object rhythm data for identifying a sleep rhythm of the object user, object time data for identifying a sleep time of the object user, and object type data for identifying a sleep type to which the object user belongs."

[0011] An information processing device involved in one embodiment can be "configured to include at least one processor, which obtains object behavior data for identifying a historical record of behavior performed by an object user, and outputs object sleep state data for identifying a sleep state of the object user from the inference model by inputting the object behavior data into an inference model generated by performing supervised learning, wherein the object sleep state data includes at least one of object property data for identifying the sleep property of the object user, object rhythm data for identifying the sleep rhythm of the object user, object time data for identifying the sleep time of the object user, and object type data for identifying the sleep type to which the object user belongs."

[0012] The method involved in one embodiment may be "a method performed by at least one processor that executes commands that can be read by a computer, comprising the following stages: the at least one processor obtains object behavior data for identifying a historical record of behaviors performed by an object user by executing the commands; and by inputting the object behavior data into an inference model generated by performing supervised learning, outputting object sleep state data for identifying the sleep state of the object user from the inference model, the object sleep state data including at least one of object property data for identifying the sleep property of the object user, object rhythm data for identifying the sleep rhythm of the object user, object time data for identifying the sleep time of the object user, and object type data for identifying the sleep type to which the object user belongs."

[0013] The method involved in another embodiment is capable of "being a method executed by at least one processor that executes commands that can be read by a computer, comprising the following stages: an acquisition stage, in which the at least one processor acquires multiple sets of teaching data by executing the commands, each set of teaching data of the multiple sets of teaching data including sample behavior data for identifying a historical record of behaviors performed by a sample user corresponding to the group and sample sleep state data for identifying a sleep state of a sample user corresponding to the group; and a generation stage, in which an inference model is generated by inputting the multiple sets of teaching data into a learning model and causing it to learn, the inference model being configured to receive object behavior data for identifying a historical record of behaviors performed by an object user and output object sleep state data for identifying the sleep state of the object user, the object sleep state data including at least one of object property data for identifying the sleep property of the object user, object rhythm data for identifying the sleep rhythm of the object user, object time data for identifying the sleep time of the object user, and object type data for identifying the sleep type to which the object user belongs."

[0014] The computer program involved in another embodiment is capable of "being executed by at least one processor to cause the at least one processor to perform the following functions: obtaining object behavior data for identifying a historical record of behaviors performed by an object user, and outputting object menopausal state data for identifying a state of menopausal symptoms of the object user from the inference model by inputting the object behavior data into an inference model generated by performing supervised learning, the object menopausal state data including at least one of object comprehensive state data for identifying a comprehensive state of the menopausal symptoms of the object user, object first state data for identifying a state of a first symptom of the object user, and object second state data for identifying a state of a second symptom of the object user."

[0015] The information processing device involved in another embodiment is capable of "being configured to have at least one processor, which obtains object behavior data for identifying the historical record of behavior performed by the object user, and outputs object menopausal state data for identifying the state of menopausal symptoms of the object user from the inference model by inputting the object behavior data into an inference model generated by performing supervised learning, wherein the object menopausal state data includes at least one of object comprehensive state data for identifying the comprehensive state of the menopausal symptoms of the object user, object first state data for identifying the state of a first symptom of the object user, and object second state data for identifying the state of a second symptom of the object user."

[0016] The method involved in another embodiment is capable of "being a method performed by at least one processor that executes commands that can be read by a computer, comprising the following stages: the at least one processor obtains object behavior data for identifying a historical record of behaviors performed by an object user by executing the commands; and by inputting the object behavior data into an inference model generated by performing supervised learning, outputting object menopausal state data for identifying the state of menopausal symptoms of the object user from the inference model, the object menopausal state data including at least one of object comprehensive state data for identifying a comprehensive state of menopausal symptoms of the object user, object first state data for identifying a state of a first symptom of the object user, and object second state data for identifying a state of a second symptom of the object user."

[0017] The method involved in another embodiment is capable of "being a method executed by at least one processor that executes commands that can be read by a computer, comprising the following stages: an acquisition stage, in which the at least one processor acquires multiple sets of teaching data by executing the commands, each set of teaching data comprising sample behavior data for identifying a historical record of behaviors performed by a sample user corresponding to the group and sample menopausal status data for identifying a state of menopausal symptoms of a sample user corresponding to the group; and a generation stage, in which an inference model is generated by inputting the multiple sets of teaching data into a learning model and causing it to learn, the inference model being configured to be input with object behavior data for identifying a historical record of behaviors performed by an object user, and outputting object menopausal status data for identifying a state of menopausal symptoms of the object user, the object menopausal status data comprising at least one of object comprehensive status data for identifying a comprehensive state of menopausal symptoms of the object user, object first status data for identifying a state of a first symptom of the object user, and object second status data for identifying a state of a second symptom of the object user".

[0018] Another embodiment involves a computer program that is capable of "enabling an information processing device to perform the following processing: obtaining first purchase data representing the purchase history of a target customer; predictive processing, using a health prediction model related to multiple health symptoms from the first to the Nth (N is an integer greater than 2) types to predict the extent of the multiple health symptoms based on the first purchase data; and providing health information indicating the predicted extent of the multiple health symptoms to the target customer, wherein the health prediction model related to the kth (1≤k≤N) health symptom calculates a numerical value indicating the extent of the kth health symptom based on the first purchase data related to the product category."

[0019] The computer program involved in another embodiment is also capable of "enabling the information processing device to further perform processing for determining a commodity category to which at least one commodity belongs based on the first purchase data, and the prediction processing uses a health prediction model related to multiple health symptoms from the first to Nth (N is an integer greater than 2) types to predict the degree of the multiple health symptoms based on the first purchase data related to the commodity category."

[0020] The computer program involved in another embodiment is also capable of "using second purchase data representing the purchase history of each of multiple subjects as an explanatory variable, and health data representing the degree of health symptoms collected from each of the multiple subjects as a target variable, and constructing a health prediction model related to the kth (1≤k≤N) health symptom through machine learning."

[0021] The computer program involved in another embodiment may also be "the machine learning is machine learning based on logistic regression."

[0022] The computer program according to another embodiment may be characterized in that "the logistic regression uses L1 normalization method and / or L2 normalization method".

[0023] The computer program involved in another method is also capable of "performing the following processing: the kth (1≤k≤N) health symptom is classified into multiple levels according to the degree of the health symptom, and each of the multiple levels is pre-associated with a numerical range corresponding to the degree of the health symptom. In the prediction processing, the level corresponding to the numerical value representing the degree of the health symptom of the target customer is determined based on the numerical range to which the numerical value belongs, thereby predicting the degree of the health symptom of the target customer."

[0024] The computer program according to another aspect may be characterized in that “the plurality of levels are three or more”.

[0025] The computer program involved in another embodiment is also capable of "enabling the information processing device to perform recommendation processing, in which, by referring to a table in which the multiple levels are respectively associated with services and / or products suitable for health symptoms at each level, recommendations related to services and / or products suitable for health symptoms at the determined level are made while providing the health information."

[0026] The computer program according to another aspect is further capable of "in the recommendation process, recommending insurance products associated with the health information by referring to a table in which the plurality of levels are associated with services suitable for health symptoms of each level."

[0027] The computer program involved in another embodiment is also capable of "introducing to the target customer a medical institution that performs preventive diagnosis and / or treatment associated with the health information by referring to a table obtained by associating each of the multiple levels with services suitable for health symptoms of each level during the recommendation processing."

[0028] The computer program involved in another embodiment can also be "the first purchase data is a purchase history of food and / or daily necessities purchased within a certain period of time, and the degree of the kth (1≤k≤N) health symptom is determined by the simplified menopausal index (SMI), the menopausal disorder evaluation scale (the Menopause Rating Scale), the Kupperman index, the Green Climacteric Scale, the premenstrual symptoms screening tool (PSST), MRS (Menopause rating Scale), WHQ (The Women's Health Questionnaire), VAS (Visual analogue scale), HFRDI (Hot Flash Related Daily Interference Scale), HFCS (Hot Flash Composite Score), MENQOL (Menopause-Specific Quality of life: Menopausal Quality of Life Scale), and Japanese Women's Menopausal Symptom Evaluation Scale.

[0029] The computer program involved in another embodiment may also be "the health symptoms of the target customer are one or more symptoms selected from menopausal symptoms and premenstrual syndrome (PMS, PMDD) symptoms, and the health prediction models related to the multiple health symptoms include one or more prediction models among the menopausal prediction model and the premenstrual syndrome (PMS, PMDD) prediction model."

[0030] The computer program according to another aspect may be such that “the degree of the k-th (1≤k≤N) health symptom is represented by one or more of a deterioration of the sleep state, a disorder of the sleep rhythm, and an inappropriate sleep time”.

[0031] The computer program involved in another embodiment may also be "the health prediction models related to the multiple health symptoms include a sleep state prediction model."

[0032] The computer program according to another embodiment may further be such that “the degree of the k-th (1≤k≤N) health symptom is represented by one or more of insufficient exercise, obesity, and a change in body weight over a certain period of time”.

[0033] The computer program involved in another embodiment is also capable of "including one or more information of whether or not each of the multiple subjects smokes and the seating behavior as an explanatory variable to perform the machine learning."

[0034] The computer program according to another embodiment may be such that “the health prediction models related to the plurality of health symptoms include a lifestyle disease prediction model related to exercise status”.

[0035] The computer program according to another embodiment may be such that "the degree of the k-th (1≤k≤N) health symptom is expressed by one or more of serum sodium, BUN (urea nitrogen) / creatinine ratio, urine osmotic pressure, urine color, and urine specific gravity."

[0036] The computer program involved in another embodiment may also be capable of "contains one or more information of serum sodium, BUN / creatinine ratio, urine osmotic pressure, urine specific gravity, water intake, alcohol intake, physical activity and urine color of each of the multiple subjects as explanatory variables to perform the machine learning."

[0037] The computer program according to another embodiment may be characterized as follows: “the health prediction models related to the plurality of health symptoms include a hydration status prediction model related to hydration status.”

[0038] The computer program involved in another embodiment may also be "the degree of the kth (1≤k≤N) health symptom is represented by any one or more of susceptibility to colds, frequency of colds, SIgA concentration, allergy symptoms, oral environment and immune status, exercise status, stress status, severe sleep symptoms, moderate sleep symptoms, mild sleep symptoms and cold symptoms."

[0039] The computer program involved in another embodiment may also be "the health prediction models related to the plurality of health symptoms include an immune status prediction model related to immune status."

[0040] Another embodiment of the computer program may be such that "the degree of the kth (1≤k≤N) health symptom is represented by the presence or absence of subjective symptoms caused by decreased nutrition, the diversity score of food intake (DVS: Dvariety Score), the shortened version of the appetite scale (SNAQ: Simplified nutritional appetite questionnaire) and any one or more of BMI (Body Mass Index)".

[0041] The computer program involved in another embodiment may also be "the health prediction models related to the plurality of health symptoms include a nutritional status prediction model related to nutritional status."

[0042] The information processing device involved in another embodiment is capable of "having: an input unit, which obtains first purchase data representing the purchase history of a target customer; a prediction unit, which predicts the degree of multiple health symptoms from the first to the Nth (N is an integer greater than 2) health symptoms based on the first purchase data; and an output unit, which provides the target customer with health information indicating the predicted degrees of the multiple health symptoms, wherein the health prediction model related to the kth (1≤k≤N) health symptom calculates a numerical value indicating the degree of the kth health symptom based on the first purchase data related to the product category."

[0043] The information processing method involved in another embodiment is capable of "being an information processing method in which an information processing device performs the following processing: obtaining first purchase data representing the purchase history of a target customer; predicting the extent of the multiple health symptoms using a health prediction model related to multiple health symptoms from the 1st to the Nth (N is an integer greater than 2) types based on the first purchase data; and providing health information indicating the predicted extent of the multiple health symptoms to the target customer, wherein the health prediction model related to the kth (1≤k≤N) health symptom of the information processing device calculates a numerical value indicating the extent of the kth health symptom based on the first purchase data related to the product category." BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a block diagram showing an example of the structure of a communication system involved in one embodiment.

[0045] Figure 2 It is shown Figure 1 2 is a block diagram of an example of a hardware configuration of the server device 10 shown.

[0046] Figure 3 It is shown by Figure 1 A flowchart of an example of operations performed by the communication system 1 shown.

[0047] Figure 4 This is a diagram showing an example of the Athens Insomnia Scale.

[0048] Figure 5A This is a diagram showing an example of the Pittsburgh Sleep Questionnaire.

[0049] Figure 5B This is a diagram showing an example of the Pittsburgh Sleep Questionnaire.

[0050] Figure 5C This is a diagram showing an example of the Pittsburgh Sleep Questionnaire.

[0051] Fig. 6A : is a figure which shows an example of a three-dimensional sleep scale.

[0052] Figure 6B : is a figure which shows an example of a three-dimensional sleep scale.

[0053] Figure 7 This is a diagram showing an example of an insomnia severity questionnaire.

[0054] Fig. 8A This is a diagram showing an example of the Munich time-based questionnaire.

[0055] Figure 8B This is a diagram showing an example of the Munich time-based questionnaire.

[0056] Fig. 9 It is shown in Figure 1 FIG. 2 is a diagram showing an example of sleep pattern data used in the communication system 1 shown.

[0057] Fig.10 It is shown in Figure 1 FIG. 2 is a diagram showing another example of sleep pattern data used in the communication system 1 shown.

[0058] Fig.11 is conceptually shown in Figure 1 FIG. 2 is a diagram showing an example of a search table used in the communication system 1 shown.

[0059] Fig.12 It is shown by Figure 1 A flowchart of another example of the operation performed by the communication system 1 shown.

[0060] Fig.13 This is a diagram showing an example of the simplified menopausal index (SMI).

[0061] Fig.14 is conceptually shown in Figure 1 FIG. 2 is a diagram showing another example of a search table used in the communication system 1 shown.

[0062] Fig.15 This is a diagram schematically showing the flow of data for performing health prediction of a customer using the health prediction system according to the embodiment.

[0063] Fig.16 1 is a diagram showing a specific example of the functional configuration of the comprehensive health prediction program 100 according to the embodiment.

[0064] Fig.17 This is a diagram showing a specific example of purchase data according to the embodiment.

[0065] Fig.18 This is a diagram showing a specific example of purchase data according to the embodiment.

[0066] Fig.19 This is a diagram showing a specific example of a table for implementing advice to customers in accordance with output values ​​of a prediction model.

[0067] Fig. 20 This is a diagram showing a specific example of a table for implementing advice to customers in accordance with output values ​​of a prediction model.

[0068] Fig.21 It is a diagram showing the process flow of the comprehensive health prediction program involved in the embodiment.

[0069] Fig. 22 This is a diagram showing a specific example of the hardware configuration of an information processing device that executes a comprehensive health prediction program according to an embodiment.

[0070] Fig.23 It is a diagram showing a specific example of the functional configuration of a learning system including a model learning program according to an embodiment.

[0071] Fig.24 This is a diagram showing a specific example of product categories.

[0072] Fig.25 This is a diagram showing a specific example of the relationship between menopausal symptoms and health issues found from the purchased items of the subjects in the H layer related to menopausal symptoms.

[0073] Fig.26 This is a diagram showing a specific example of the product category for women with menopausal symptoms belonging to the H layer.

[0074] Fig. 27 It is a diagram showing the process flow of the model learning program involved in the embodiment.

[0075] Fig.28 This is a diagram showing a specific example of the hardware configuration of an information processing device that executes a model learning program according to an embodiment. DETAILED DESCRIPTION

[0076] Various embodiments of the present invention are described below with reference to the accompanying drawings. In addition, the same reference numerals are used to mark the same components in the accompanying drawings. In addition, it is hoped that the components shown in a certain drawing are sometimes omitted in other drawings for the sake of convenience. In addition, it is also hoped that the attached drawings are not necessarily recorded in the correct scale.

[0077] It should not be interpreted that the various systems, methods and devices described in this specification are limited by any method. In fact, the present disclosure is directed to each of the various embodiments disclosed, the embodiments obtained by combining these various embodiments, and all new features and methods in the embodiments obtained by combining parts of these various embodiments. The various systems, methods and devices described in this specification are not limited to a specific method, a specific feature, or a combination of such a specific method and a specific feature, and the things and methods described in this specification are not required to have one or more specific effects or to solve the problem. In addition, the various features or methods in the various embodiments described in this specification, or a part of such features or methods, can be used in combination with each other.

[0078] For convenience, the actions of several methods disclosed in this specification are recorded in a specific order, but it should be understood that as long as a specific order is not required by a specific article below, the recording of this method includes the situation of rearranging the above order of actions. For example, when there are several actions recorded in sequence, these actions can be rearranged or performed simultaneously. In addition, for the purpose of simplicity, the drawings do not show various methods that can use various matters and methods recorded in this specification together with other matters and methods.

[0079] The description of the theory of action, scientific principle or other theory presented in this specification in association with the device or method of the present disclosure is provided for better understanding and is not intended to limit the scope of protection. The device and method in the attached claims are not limited to the device and method that operates by the method based on such theory of action.

[0080] The various methods disclosed in this specification can be implemented using multiple commands that are stored in one or more media that can be read by a computer and can be executed by a computer, and can be executed in a computer. The one or more media can be, for example, a non-transitory computer-readable storage medium such as at least one optical media disk, multiple volatile memory components, or multiple non-volatile memory components. Here, the multiple volatile memory components include, for example, DRAM or SRAM. In addition, the multiple non-volatile memory components include, for example, hard disk drives and solid-state drives (SSDs). In addition, the computer includes, for example, any computer that can be obtained on the market, including smartphones and other mobile devices with hardware for computing.

[0081] Such multiple commands that can be executed by a computer for implementing the technology disclosed in this specification can be stored in one or more computer-readable media (e.g., non-transitory computer-readable storage media) together with any data generated and used during the implementation of the various embodiments disclosed in this specification. Such multiple commands that can be executed by a computer can, for example, be part of an independent software application, or can be part of a software application that is accessed or downloaded via a web browser or other software application (such as a remote computing application). Such software can, for example, be executed in a single local computer (e.g., as a process that can be executed in any preferred computer available on the market), or in a network environment (e.g., the Internet, a wide area network, a local area network, a client-server network (such as a cloud computing network) or other such network) using one or more network computers.

[0082] For the sake of clarity, only specific selected various methods among various software-based implementations are recorded. Other detailed matters known in the art are omitted. For example, the technology disclosed in this specification is not limited to a specific computer language or program. For example, the technology disclosed in this specification can be executed by software described by C, C++, Java (registered trademark), or any other preferred programming language. Similarly, the technology disclosed in this specification is not limited to a specific computer or a specific type of hardware. The specific details of the preferred computer and hardware are well known and do not need to be described in detail in this specification.

[0083] Furthermore, various implementations based on such software (e.g., including a plurality of commands executable by a computer for causing a computer to perform any of the various methods disclosed in this specification) can be uploaded, downloaded, or accessed remotely through preferred communication means. Such preferred communication means include, for example, the Internet, the World Wide Web, an intranet, a software application, cables (including fiber optic cables), magnetic communications, electromagnetic communications (including RF communications, microwave communications, infrared communications), electronic communications, or other such communication means.

[0084] I. Chapter 1

[0085] 1. Summary

[0086] In the technology disclosed in the present application, in short, at least one information processing device obtains object behavior data for identifying the historical records of behaviors performed by a user (object user) set as an inference object, and inputs the object behavior data into an inference model, thereby being able to output data (health status data) for identifying the health status of the above-mentioned object user from the above-mentioned inference model.

[0087] In addition, at least one information processing device can also determine at least one target product and / or at least one target service corresponding to the health status data of the target user among multiple products and / or multiple services. And at least one information processing device can also output information for identifying at least one target service and / or at least one target service determined in this way.

[0088] 2. Communication system structure

[0089] This technique can be used Figure 1 The communication system shown is implemented. Figure 1 This is a block diagram showing an example of the structure of a communication system involved in one embodiment.

[0090] like Figure 1 As shown, a communication system 1 according to one embodiment may include, for example, at least one server device (information processing device) 10 that can be connected to a communication network 2 and at least one terminal device (information processing device) 20 that can be connected to the communication network 2. Figure 1 In the example of at least one server device 10, two server devices 10A and 10B are shown, but any number of server devices 10 can be used. Figure 1 In the embodiment, two terminal devices 20A and 20B are exemplified as the at least one terminal device 20, but any number of server devices 10 can be used.

[0091] Each server device 10 can be connected to at least one other server device 10 and / or at least one terminal device 20 via the communication network 2. In addition, each terminal device 20 can be connected to at least one other terminal device 20 and / or at least one server device 10 via the communication network 2.

[0092] Each server device 10 can be any information processing device. Such information processing devices can include, but are not limited to, personal computers, workstations, supercomputers, and mainframes. Each terminal device 20 can be any information processing device. Such information processing devices can include, but are not limited to, mobile phones, smart phones, tablet computers, personal computers, workstations, and portable information terminals.

[0093] In one example, at least one server device 10 can be an information processing device managed by an operating company that operates the following services: a service for inferring the health status of any user ("inference service"), and / or a service for recommending at least one target product, etc., corresponding to the health status of any user ("recommendation service"). In addition, for convenience, the term "inference service, etc." is sometimes used as a term to refer to the inference service and / or the recommendation service.

[0094] In one example, at least one server device 10 and / or at least one terminal device 20 may be an information processing device managed by a business user and / or an individual user who receives provision of an inference service or the like from the above-mentioned operating company.

[0095] The above-mentioned action of acquiring the object behavior data can be performed by at least one server device 10 and / or at least one terminal device 20 .

[0096] The above-mentioned inference model can be generated by at least one server device 10 and / or at least one terminal device 20. The inference model generated in this way can be maintained by at least one server device 10 and / or at least one terminal device 20. The inference model maintained in this way can be used by at least one server device 10 and / or at least one terminal device 20.

[0097] The above-mentioned action of outputting health status data from the inference model can be performed by at least one server device 10 and / or at least one terminal device 20 .

[0098] The above-mentioned action of determining at least one target commodity and / or at least one target service can be performed by at least one server device 10 and / or at least one terminal device 20. The above-mentioned action of outputting data for identifying at least one target service and / or at least one target service can be performed by at least one server device 10 and / or at least one terminal device 20.

[0099] In addition, the communication network 2 can include but is not limited to a mobile phone network, a wireless network, a fixed telephone network, the Internet, an intranet, a local area network (LAN), a wide area network (WAN) and / or an Ethernet (registered trademark) network. The above wireless network can include, for example, but is not limited to, an RF connection via Bluetooth (registered trademark), WiFi (such as IEEE802.11a / b / n), WiMax, cellular, satellite, laser, infrared.

[0100] 3. Hardware structure of each information processing device

[0101] Next, an example of the hardware configuration of each information processing device (server device 10, terminal device 20) will be described.

[0102] (1) Hardware Configuration of Server Device 10

[0103] Figure 2 It is shown Figure 1 A block diagram of an example of a hardware configuration of the server device 10 shown in FIG. Figure 2 In the figure, the reference numerals in brackets are recorded in association with the terminal device 20 as described later).

[0104] like Figure 2 As shown, the server device 10 can include a central processing unit 11, a main storage device 12, an input / output interface device 13, an input device 14, an auxiliary storage device 15, and an output device 16. These devices are connected to each other via a data bus and / or a control bus.

[0105] The central processing unit 11 is called a "CPU" and can perform operations on commands and data stored in the main storage device 12 and store the operation results in the main storage device 12. In addition, the central processing unit 11 can control the input device 14, the auxiliary storage device 15, and the output device 16 through the input and output interface device 13. The server device 10 can include one or more central processing units 11.

[0106] The main storage device 12 is called "memory", and can store commands and data received from the input device 14, the auxiliary storage device 15, and the communication network 2 (terminal device 20, etc.) via the input / output interface device 13, as well as the calculation results of the central processing unit 11. The main storage device 12 can include, but is not limited to, computer-readable media such as volatile memory (e.g., registers, caches, random access memories (RAM)), non-volatile memory (e.g., read-only memory (ROM), EEPROM, flash memory), and storage (e.g., hard disk drive (HDD), solid state drive (SSD), magnetic tape, optical media). For ease of understanding, the term "computer-readable storage medium" can include media for data storage such as memory and storage, but does not include transmission media such as for transmitting modulated data signals, i.e., temporary signals.

[0107] The auxiliary storage device 15 is a storage device having a larger capacity than the main storage device 12. The auxiliary storage device 15 stores commands and data (computer programs) constituting specific application programs in advance, and can send these commands and data (computer programs) to the main storage device 12 via the input / output interface device 13 by being controlled by the central processing unit 11. The auxiliary storage device 15 can include, but is not limited to, a magnetic disk device and / or an optical disk device.

[0108] The specific applications mentioned here can include but are not limited to operating systems, various software applications (including web browsers), dedicated applications executed to provide inference services, etc., dedicated applications executed to receive inference services, etc.

[0109] The input device 14 is a device for taking in data from the outside, and may include, but is not limited to, a keyboard, a touch panel, a button, a mouse, and / or a sensor (microphone, camera), etc.

[0110] The output device 16 can include, but is not limited to, a display device, a touch panel, and / or a printer device.

[0111] In such a hardware structure, the central processing unit 11 can sequentially load the commands and data (computer program) constituting the above-mentioned specific application program stored in the auxiliary storage device 15 into the main storage device 12, and perform operations on the loaded commands and data. Thus, the central processing unit 11 can control the output device 16 via the input / output interface device 13, or transmit and receive various information (data) with other devices (such as other server devices 10 and / or terminal devices 20, etc.) via the input / output interface device 13 and the communication network 2.

[0112] Thus, the server device 10 can execute the operations related to the operations of providing the inference service and / or the operations of receiving the provision of the inference service by executing the installed specific application program (see Figure 3 and Fig.12 In addition to or instead of this, the server device 10 can also access other server devices 10 by executing a browser to perform actions related to providing an inference service and / or receiving an inference service (see Figure 3 and Fig.12 etc. (various actions described later).

[0113] In addition, the server device 10 can also include one or more microprocessors and / or graphic processing units (GPUs) instead of the central processing unit 11, or include one or more microprocessors and / or graphic processing units (GPUs) while including the central processing unit 11.

[0114] (2) Hardware Configuration of Terminal Device 20

[0115] like Figure 2 As shown in the brackets in , the terminal device 20 can have substantially the same hardware structure as the server device 10. Figure 2 As shown, the terminal device 20 can include a central processing unit 21, a main storage device 22, an input / output interface device 23, an input device 24, an auxiliary storage device 25, and an output device 26. These devices are connected to each other via a data bus and / or a control bus. Each of these devices is as described in connection with the server device 10.

[0116] The specific application programs stored in the auxiliary storage device 25 and executed by the central processing unit 21 may include, but are not limited to, operating systems, various software applications (including web browsers), and dedicated applications executed to receive provision of inference services, etc. In addition, the specific application programs may also include dedicated applications executed to provide inference services, etc.

[0117] In this hardware structure, the central processing unit 21 can sequentially load the commands and data (computer programs) constituting the above-mentioned specific application programs stored in the auxiliary storage device 25 into the main storage device 22, and perform operations on the loaded commands and data. Thus, the central processing unit 21 can control the output device 26 via the input / output interface device 23, or transmit and receive various information (data) with other devices (such as the server device 10 and / or other terminal devices 20) via the input / output interface device 23 and the communication network 2.

[0118] Thus, the terminal device 20 can execute the above-mentioned installed specific application program to perform actions related to the action of receiving provision of the inference service and / or the action of providing the inference service and the like (see Figure 3 and Fig.12 In addition to or instead of this, the server device 10 can also access other server devices 10 by executing a browser to perform actions related to providing an inference service and / or receiving an inference service (see Figure 3 and Fig.12 etc. (various actions described later).

[0119] In addition, the terminal device 20 can also include one or more microprocessors and / or graphic processing units (GPUs) instead of the central processing unit 21, or include one or more microprocessors and / or graphic processing units (GPUs) while including the central processing unit 21.

[0120] 4. Actions performed by the communication system 2 to provide sleep state data related to the sleep state

[0121] Next, refer to Figure 3 A specific example of the operation performed by the communication system 2 to provide the sleep state data related to the sleep state will be further described. Figure 3 It is shown by Figure 1 A flowchart of an example of operations performed by the communication system 1 shown.

[0122] (1) Step 1000

[0123] First, in step (hereinafter referred to as “ST”) 1000 , a server device 10A as an information processing device managed by, for example, a company operating an inference service or the like can acquire and store a plurality of sets of teaching data for generating an inference model.

[0124] Each set of teaching data constituting the plurality of sets of teaching data may include data for identifying a history of behaviors performed by a sample user corresponding to the group ("sample behavior data") and data for identifying a sleep state of a sample user corresponding to the group ("sample sleep state data"). For example, the first set of teaching data may include sample behavior data for identifying a history of behaviors performed by a sample user (A) corresponding to the first group and sample sleep state data for identifying A's sleep state, and the second set of teaching data may include sample behavior data for identifying a history of behaviors performed by a sample user (B) corresponding to the second group and sample sleep state data for identifying B's sleep state.

[0125] (1A) Sample behavior data

[0126] Each sample behavior data may be data for identifying at least one (one or more) of a plurality of predetermined behaviors. Here, each of the plurality of predetermined behaviors may be a behavior that may at least partially exert a favorable or unfavorable influence on the sleep state of a person who performs the behavior. Such a plurality of predetermined behaviors may include, but are not limited to, the behaviors exemplified below.

[0127] (Examples of behaviors that may at least partially exert a favorable influence on a person's sleep state)

[0128] Performed aerobic exercise for at least one hour at a moderate pace 2 or more days in a week.

[0129] Get up at a fixed time every day.

[0130] Limit caffeinated beverages to 2 cups or less per day.

[0131] · The working hours per day are less than 8 hours.

[0132] Purchased more than 10 cups of coffee (this behavior can be identified, for example, by data including the product category to which coffee belongs and the quantity purchased).

[0133] (Examples of behaviors that may at least partially exert adverse effects on a person's sleep state)

[0134] Performed aerobic physical activity for at least 1 hour at a pace no more than 1 day per week.

[0135] Wake up at a different time each day.

[0136] Drink more than 3 cups of caffeinated beverages per day.

[0137] · The working hours per day are more than 13 hours.

[0138] · Purchased more than 5 cups of milk (this behavior can be identified, for example, by data including the product category to which the milk belongs and the quantity purchased).

[0139] Each sample behavior data can be data for identifying at least one behavior among a plurality of predetermined behaviors exemplified in this manner.

[0140] In one example, the goods purchased by the user and / or the services used by the user may have a tendency to exert a favorable or unfavorable influence on the sleep state of the user, so the predetermined multiple behaviors may also include goods purchased by the sample user (or the category to which the goods belong) and services used by the sample user (or the category to which the services belong). Such behaviors can be generated, for example, using POS (Point Of Sales) data owned by retail stores (drug stores, supermarkets, convenience stores, etc.).

[0141] In one example, a plurality of predetermined behaviors are each assigned unique identification data (such as English letters, numbers, or a combination thereof), and each sample behavior data may include such identification data. For example, when using POS data, each of the plurality of sample behavior data may include, but is not limited to, JICFS classification (classification code) corresponding to the purchased commodity, commodity category, purchase amount, purchase quantity, purchase frequency, gender, age, and / or purchase date, etc.

[0142] The server device 10A as an information processing device managed by, for example, a company operating an inference service or the like can acquire such sample behavior data by, for example, at least one of the following exemplified methods.

[0143] The information processing device receives POS data via the communication network 2 from at least one other server device 10 (for example, another server device 10 installed in a drug store or the like for storing POS data, or another server device 10 capable of accessing the POS data).

[0144] The information processing device receives an email or a web page recording answers to a predetermined questionnaire (questionnaire related to behavior performed by the sample user) from at least one terminal device 20 and / or at least one server device 10 via the communication network 2 .

[0145] The information processing device receives the content of the answers to the above-mentioned predetermined questionnaire from at least one server device 10 that has collected the answers from a plurality of users via the communication network 2 .

[0146] The information processing device receives the contents of the answers to the above-mentioned predetermined questionnaire from a plurality of users via a recording medium (USB memory, DVD-ROM, etc.) on which the answers are collected.

[0147] (1B) Sample sleep state data

[0148] The sample sleep state data included in each set of teaching data may include, but is not limited to, at least one of the data exemplified below.

[0149] Data for identifying the sleep properties of a sample user corresponding to the group ("sample property data")

[0150] Data for identifying the sleep rhythm of a sample user corresponding to the group ("sample rhythm data")

[0151] Data for identifying the sleep time of a sample user corresponding to the group ("sample time data")

[0152] Data for identifying the sleep type to which (the sleep of) a sample user corresponding to the group belongs ("sample sleep type data")

[0153] First, the sample property data may be data for identifying whether the sleep quality of the sample user is good (e.g., to what extent it is good) or bad (e.g., to what extent it is bad). In one example, the sample property data may be data for identifying a score obtained by comprehensively evaluating the sleep quality of the sample user. In this case, the sleep quality of the sample user may be data for identifying a score calculated using any of the methods exemplified below.

[0154] Based on the sample users' responses to the Athens Insomnia Scale (AIS) (ref. Figure 4 ) is a score calculated using the method described in the Athens Insomnia Scale, or a corrected score obtained by performing an operation (addition, subtraction, multiplication, division, etc.) on the score using any method.

[0155] Based on the Pittsburgh Sleep Questionnaire (PSQI) (reference FIG. 5A to FIG. 5C ) is calculated using the method described in the Pittsburgh Sleep Questionnaire, or a corrected score obtained by performing an operation (addition, subtraction, multiplication, division, etc.) on the score using any method.

[0156] Based on the sample users' 3D Sleep Scale (3DSS) (refer to Fig. 6A and Figure 6B ) is calculated using the method described in the three-dimensional sleep scale, or a corrected score obtained by performing an operation (addition, subtraction, multiplication, division, etc.) on the score using any method.

[0157] Based on the Insomnia Severity Questionnaire (ISI) (reference Figure 7 ) is calculated using the method described in the Insomnia Severity Questionnaire, or a corrected score obtained by performing an operation (addition, subtraction, multiplication, division, etc.) on the score using any method.

[0158] A score calculated based on the sample user's answers to questions in any other questionnaire or scale using the method described in the questionnaire or scale, or a modified score calculated by calculating the score using any method

[0159] Second, the sample rhythm data can be data for identifying whether the quality of the sample user's rhythm is good (e.g., to what extent it is good) or bad (e.g., to what extent it is bad). In one example, the sample rhythm data can be data for identifying a score obtained by evaluating the sample user's sleep rhythm.

[0160] In this case, the sleep rhythm of the sample user can be, for example, data for identifying the absolute value of the difference between the central time of the sample user's usual sleep time period and the central time of the sleep time period on holidays, that is, social jet lag (or a value obtained by calculating the social jet lag using any method). Here, for example, when the usual sleep time period of a sample user is from 0 am to 6 am, the central time is 3 am, and when the sleep time period on holidays of the sample user is from 3 am to 11 am, the central time is 7 am, so the social jet lag is 4 hours. Generally speaking, the smaller (or larger) the social jet lag, the better (or worse) the sleep rhythm of the sample user.

[0161] In one example, the sleep time periods of the sample users on weekdays and weekends can be obtained from the Munich Chronotype Questionnaire (MCTQ) (see Fig. 8A and Figure 8B ) is extracted from the answer content of the question items recorded in, and in another example, can be obtained from the sample user through email or web page.

[0162] Third, the sample time data can be data for identifying the sleep time of the sample user (sleep time for one day). In one example, the sleep time of the sample user can be obtained from the Munich Chronotype Questionnaire (MCTQ) (refer to Fig. 8A and Figure 8B) is extracted from the answer content of the question items recorded in, and in another example, can be obtained from the sample user through email or web page.

[0163] Fourth, the sample sleep type data can be data for identifying the sleep type of the sample user (sleep). As an example, the sample sleep type data can be data for identifying the type of the sample user (sleep) determined based on the sample property data of the sample user and the sample rhythm data of the sample user. In this case, the sleep type can be determined, for example, by the following method.

[0164] Fig. 9 It is shown in Figure 1 FIG. 1 is a diagram showing an example of sleep type data used in the communication system 1 shown in FIG. Fig. 9 As shown, the score identified by the sample property data can be configured on one of the vertical axis and the horizontal axis (here the horizontal axis), and the value identified by the sample rhythm data can be configured on the other of the vertical axis and the horizontal axis (here the vertical axis). On the above-mentioned one axis, at least one (here one) threshold value, namely, threshold value A1 (6 points) can be set for the score identified by the sample property data. On the above-mentioned other axis, at least one (here one) threshold value, namely, threshold value B1 (1 hour) can be set for the value identified by the sample rhythm data. Thus, four different sleep types (here sleep types 10, 11, 20, and 21) divided from the perspective of sleep properties and sleep rhythm can be formed based on threshold value A1 and threshold value B1. In this example, it can be said that sleep type 10 is a type with a high probability of causing the most significant obstacles during sleep, sleep type 21 is a type with a high probability of causing the least obstacles during sleep, and sleep types 11 and 20 are types located between sleep types 10 and 21. The sleep type data can be data for identifying a sleep type among these four sleep types.

[0165] Fig.10 It is shown in Figure 1 FIG. 1 is a diagram showing another example of sleep type data used in the communication system 1 shown in FIG. Fig.10In the example shown, on one axis, for example, two thresholds, namely, threshold A1 (6 points) and threshold A2 (8 points), can be set for the score identified by the sample property data. On the other axis, for example, two thresholds, namely, threshold B1 (1 hour) and threshold B2 (3 hours), can be set for the value identified by the sample rhythm data. Thus, nine different sleep types (here, sleep types 100, 101, 102, 200, 201, 202, 300, 301, 302) divided from the viewpoint of sleep properties and sleep rhythm can be formed according to thresholds A1, A2 and thresholds B1, B2. In this example, it can be said that sleep type 300 is a type with a high probability of causing the most significant obstacles during sleep, sleep type 202 is a type with a high probability of causing the least obstacles during sleep, and the remaining sleep types are types between sleep types 300 and 202. The sleep type data can be data for identifying a sleep type among these nine sleep types.

[0166] Fig. 9 and Fig.10 The example shown is merely an example for convenience, and any number of thresholds can be set for the score identified by the sample property data, and any number of thresholds can be set for the value identified by the sample rhythm data. By setting more thresholds for the sample property data and the sample rhythm data, more different sleep types can be generated based on the viewpoints of sleep properties and sleep rhythms.

[0167] The server device 10A (and / or the terminal device 20) as an information processing device, for example, managed by a company operating an inference service, can combine such sample behavior data with sample sleep state data for each sample user, and generate and store a set of teaching data for the sample user. To achieve this, for example, the information processing device can extract sample behavior data with identification data for identifying a certain sample user from the acquired multiple sample behavior data, extract sample sleep state data with identification data for identifying a sample user who is the same user as the sample user from the acquired multiple sample sleep state data, and combine them to generate and save a set of teaching data related to the sample user. By performing the same processing on multiple sample users, the information processing device can generate and save multiple sets of teaching data.

[0168] (2) Step 1002

[0169] Return to Figure 3Then, in ST1002, a server device 10A (and / or a terminal device 20) as an information processing device, such as a server device managed by a company operating an inference service, uses the multiple sets of teaching data acquired in ST1000 to make the learning model learn through machine learning, thereby generating an inference model. The machine learning mentioned here can include but is not limited to any known machine learning, such as a gradient boosting decision tree (such as LightGBM), a neural network, or a linear regression.

[0170] The information processing device can perform learning by sequentially inputting a plurality of sets of teaching data into the learning model. Here, as described above, each set of teaching data can include sample behavior data of a sample user corresponding to the set and sample sleep state data of a sample user corresponding to the set. The information processing device can use the sample behavior data included in each set of teaching data as an explanatory variable and the sample sleep state data included in the set of teaching data as a target variable.

[0171] (2A) Example of machine learning

[0172] When the information processing device uses a gradient boosting decision tree, for example, when the sample behavior data (explanatory variables) contained in a set of teaching data is input into the learning data, a decision tree can be created and added to the learning model to improve the objective function calculated based on the sample sleep state data (target variable) and the inferred value (i.e., the value output from the learning model). The information processing device can repeat this process for the number of decision trees determined by the hyperparameter. For the second and subsequent decision trees, the difference between the target variable and the inferred value based on the decision tree created before can be learned. In this way, the information processing device can determine the weights of the branches and leaves of each decision tree. That is, in the gradient boosting decision tree, each decision tree uses the explanatory variables in the teaching data as input data and the values ​​of the leaf nodes reached by tracing back the branches of the decision tree as the data to be output (target variables). In addition, the methods of each branch of the decision tree and the values ​​of each leaf node can be optimized to reduce the error between the data (target variable) and the target variable in the teaching data.

[0173] For example, when using a neural network including an input layer, an intermediate layer, and an output layer, an information processing device can use the error back propagation method to optimize (learn) the parameters (weights) included in the neural network so that the error between the data (target variable) output from the output layer by inputting the explanatory variables in the teaching data into the input layer and the target variable in the teaching data is reduced.

[0174] When using linear regression, the information processing device can learn the parameters of the linear regression model in order to improve the error between the target variable and the inferred value. The linear regression model is divided into an input layer and an output layer, and the least square method can be used to optimize the parameters included in the linear regression model so that the error between the data (target variable) output from the output layer by inputting the explanatory variable in the teaching data as input data into the input layer and the target variable in the teaching data becomes smaller.

[0175] (2B) The first plurality of sets of teaching data included in the plurality of sets of teaching data

[0176] In one example, the information processing device can generate (update) an inference model using the first plurality of teaching data included in the plurality of teaching data. Each set of teaching data included in the first plurality of teaching data can include sample behavior data of a sample user corresponding to the set and sample sleep state data of a sample user corresponding to the set. Here, the sample sleep state data in each set of teaching data can include sample property data of a sample user corresponding to the set and sample rhythm data of the sample user.

[0177] In this case, the information processing device can perform preprocessing on each set of teaching data included in the first plurality of sets of teaching data. Specifically, the information processing device uses the sample property data and sample rhythm data included in each set of teaching data, referring to Fig. 9 and Fig.10 The above-mentioned processing is performed, thereby preprocessing such as generating sample sleep type data can be performed for the group. Thus, the information processing device can input the preprocessed first plurality of groups of teaching data (each group of teaching data includes sample behavior data of a sample user corresponding to the group and sample sleep type data of the sample user) into the learning model to generate (update) an inference model. This inference model can, as described later, respond to the input of the object behavior data of the history record for identifying the behavior performed by the object user as an explanatory variable, and output at least the object sleep type data for identifying the sleep type to which the object user belongs.

[0178] (2C) The second plurality of sets of teaching data included in the plurality of sets of teaching data

[0179] In one example, the information processing device can generate (update) an inference model using a second plurality of sets of teaching data included in the plurality of sets of teaching data. Each set of teaching data included in the second plurality of sets of teaching data can include sample behavior data of a sample user corresponding to the set and sample sleep state data of a sample user corresponding to the set. Here, the sample sleep state data in each set of teaching data can include sample sleep type data of a sample user corresponding to the set.

[0180] In this case, the information processing device can input a second plurality of groups of teaching data (each group of teaching data includes sample behavior data of a sample user corresponding to the group and sample sleep type data of the sample user) into the learning model to generate (update) an inference model. Such an inference model can output at least target sleep type data for identifying the sleep type to which the target user belongs in response to target behavior data of a history record of behaviors performed by the target user being input as an explanatory variable as described later.

[0181] (2D) The third set of teaching data included in the multiple sets of teaching data

[0182] In one example, the information processing device can generate (update) an inference model using a third plurality of sets of teaching data included in the plurality of sets of teaching data. Each set of teaching data included in the third plurality of sets of teaching data can include sample behavior data of a sample user corresponding to the set and sample sleep state data of a sample user corresponding to the set. Here, the sample sleep state data in each set of teaching data can include sample property data of a sample user corresponding to the set and sample rhythm data of the sample user.

[0183] In this case, the information processing device can input a third plurality of groups of teaching data (each group of teaching data includes sample behavior data of a sample user corresponding to the group, sample property data of the sample user, and sample rhythm data of the sample user) into the learning model to generate (update) an inference model. Such an inference model can, as described later, respond to input of object behavior data of a history record for identifying behaviors performed by the object user as an explanatory variable, output at least object property data for identifying the sleep property of the object user and object rhythm data for identifying the sleep rhythm of the object user.

[0184] (2E) the fourth plurality of teaching data included in the plurality of teaching data

[0185] In one example, the information processing device can generate (update) an inference model using a fourth plurality of sets of teaching data included in the plurality of sets of teaching data. Each set of teaching data included in the fourth plurality of sets of teaching data can include sample behavior data of a sample user corresponding to the set and sample sleep state data of a sample user corresponding to the set. Here, the sample sleep state data in each set of teaching data can include sample property data of a sample user corresponding to the set.

[0186] In this case, the information processing device can input a fourth plurality of groups of teaching data (each group of teaching data includes sample behavior data of a sample user corresponding to the group and sample property data of the sample user) into the learning model to generate (update) an inference model. Such an inference model can output at least object property data for identifying the sleep property of the object user in response to object behavior data for identifying the history of behaviors performed by the object user as an explanatory variable as described later.

[0187] (2F) The fifth plurality of teaching data included in the plurality of teaching data

[0188] In one example, the information processing device can generate (update) an inference model using the fifth plurality of teaching data included in the plurality of teaching data. Each set of teaching data included in the fifth plurality of teaching data can include sample behavior data of a sample user corresponding to the set and sample sleep state data of a sample user corresponding to the set. Here, the sample sleep state data in each set of teaching data can include sample rhythm data of a sample user corresponding to the set.

[0189] In this case, the information processing device can input a fifth plurality of groups of teaching data (each group of teaching data includes sample behavior data of a sample user corresponding to the group and sample rhythm data of the sample user) into the learning model to generate (update) an inference model. Such an inference model can output at least target rhythm data for identifying the sleep rhythm of the target user in response to target behavior data of a history record of a behavior performed by the target user being input as an explanatory variable as described later.

[0190] (2G) The sixth set of teaching data included in the multiple sets of teaching data

[0191] In one example, the information processing device can generate (update) an inference model using the fifth plurality of teaching data included in the plurality of teaching data. Each set of teaching data included in the sixth plurality of teaching data can include sample behavior data of a sample user corresponding to the set and sample sleep state data of a sample user corresponding to the set. Here, the sample sleep state data in each set of teaching data can include sample time data of a sample user corresponding to the set.

[0192] In this case, the information processing device can input the sixth plurality of groups of teaching data (each group of teaching data includes sample behavior data of a sample user corresponding to the group and sample time data of the sample user) into the learning model to generate (update) an inference model. Such an inference model can output at least object time data for identifying the sleep time of the object user in response to object behavior data for identifying the history of behaviors performed by the object user as an explanatory variable as described later.

[0193] Such generation (updating) of the inference model can be performed by another server device 10 (and / or terminal device 20) managed by another enterprise, instead of the server device 10A (and / or terminal device 20) managed by the enterprise operating the inference service, etc., or together with the server device 10A (and / or terminal device 20) managed by the enterprise operating the inference service, etc. The inference model generated (updated) by the other server device 10 (terminal device 20) can be transmitted to any server device 10 including the server device 10A, or can be accessed and used by any server device 10 including the server device 10A.

[0194] In addition, the names such as "first" to "sixth" marked in front of "multiple sets of teaching data" are used to conveniently distinguish "multiple sets of teaching data" marked with one of these names from "multiple sets of teaching data" marked with other of these names, and do not limit any order or content, etc.

[0195] (3) Step 1004

[0196] When referring again Figure 3 When, then, in ST1004, the server device 10A (and / or the terminal device 20) as an information processing device, for example, managed by a company operating an inference service, etc., can obtain target behavior data for identifying a history of behaviors performed by a target user who is the target of inference, for example, from the terminal device 20 (or other server device 10) that accesses the server device 10A by executing a specific application (an application for receiving provision of the inference service, etc.), and / or the terminal device 20 that accesses the server device 10A by executing a browser. The target behavior data can be data for identifying at least one (one or more) behaviors performed by the target user.

[0197] The acquisition of such object behavior data can be performed by other server devices 10 (and / or terminal devices 20) managed by other enterprises, instead of the server device 10A (and / or terminal device 20) managed by the enterprise operating the inference service, etc., or together with the server device 10A (and / or terminal device 20) managed by the enterprise operating the inference service, etc.

[0198] (4) Step 1006

[0199] Next, in ST1006, the server device 10A (and / or the terminal device 20) as an information processing device, for example, managed by a company operating an inference service, can obtain the target sleep state data for identifying the sleep state of the target user using the target behavior data acquired in ST1004 and the inference model generated in ST1002. Specifically, the server device 10A can output the target sleep state data as a target variable from the inference model by inputting the target behavior data as an explanatory variable into the inference model.

[0200] The target sleep state data can include at least one of the following data.

[0201] Data used to identify the sleep characteristics of the subject user ("subject characteristic data")

[0202] Data for identifying the sleep rhythm of the target user ("target rhythm data")

[0203] Data for identifying the sleep time of the target user ("target time data")

[0204] Data for identifying the sleep type to which the target user belongs (“target sleep type data”)

[0205] Which data the inference model outputs as the target sleep state data may depend on which data was used as the sample sleep state data when generating (updating) the inference model.

[0206] For example, when the first plurality of teaching data are used as sample sleep state data when generating (updating) the inference model (item "4(2)(2B)" above), the inference model can at least output the object sleep type data as the object sleep state data.

[0207] For example, when the second plurality of teaching data sets are used as sample sleep state data when generating (updating) the inference model (item "4(2)(2C)" above), the inference model can output at least the subject sleep type data as the subject sleep state data.

[0208] For example, when a third plurality of teaching data is used as sample sleep state data when generating (updating) an inference model (the above-mentioned "4(2)(2D)" item), the inference model is able to output at least object property data and object rhythm data as object sleep state data.

[0209] For example, when the fourth plurality of sets of teaching data are used as sample sleep state data when generating (updating) the inference model (the above-mentioned item "4(2)(2E)"), the inference model can output at least the object property data as the object sleep state data.

[0210] For example, when the fifth plurality of sets of teaching data are used as sample sleep state data when generating (updating) the inference model (the above-mentioned item "4(2)(2F)"), the inference model can output at least the target rhythm data as the target sleep state data.

[0211] For example, when the sixth plurality of sets of teaching data are used as sample sleep state data when generating (updating) the inference model (the above-mentioned item "4(2)(2G)"), the inference model can output at least the object time data as the object sleep state data.

[0212] For example, when the first plurality of teaching data sets are used as sample sleep state data to generate (update) the inference model (the above "4(2)(2B)" item), and then the sixth plurality of teaching data sets are used as sample sleep state data to generate (update) the inference model (the above "4(2)(2G)" item), the inference model can output at least the object sleep type data and the object time data as the object sleep state data. And, when the fifth plurality of teaching data sets are used as sample sleep state data to generate (update) the inference model (the above "4(2)(2F)" item), the inference model can output at least the object sleep type data, the object time data, and the object rhythm data as the object sleep state data.

[0213] As a method of using the estimation model, any of the methods exemplified below can be used.

[0214] The server device 10A acquires and stores the inference model generated in ST1004 , inputs the subject behavior data to the inference model, and acquires the subject sleep state data output from the inference model.

[0215] The server device 10A transmits the subject behavior data to an external device (another arbitrary server device 10 or an arbitrary terminal device 20) that acquires and stores the inference model generated in ST1004 via the communication network 2, thereby causing the inference model that has been input with the subject behavior data as an explanatory variable to output the subject sleep state data. Thereafter, the server device 10A receives the subject sleep state data output by the inference model from the external device via the communication network 2.

[0216] The acquisition of such object sleep state data can be performed by other server devices 10 (and / or terminal devices 20) managed by other enterprises, instead of or together with the server device 10A (and / or terminal device 20) managed by the enterprise operating the inference service, etc.

[0217] (5) Step 1008

[0218] Next, in ST1008, the server device 10A (and / or the terminal device 20) as an information processing device, for example, managed by a company operating an inference service, can use the target sleep state data acquired in ST1006 to determine at least one target product / one target service corresponding to the target sleep state data. Such a target product / target service can be suitable for the current (inferred) sleep state of the target user identified by the target sleep state data.

[0219] Specifically, in one example, the server device 10A can first prepare and store a search table that stores the target sleep state data in association with multiple products and / or multiple services. Next, the server device 10A can input the target sleep state data acquired in ST1006 into the search table, and can acquire data ("recommendation data") for identifying at least one target product and / or at least one target service from the search table.

[0220] Fig.11 is conceptually shown in Figure 1 FIG. 1 is a diagram showing an example of a search table used in the communication system 1 shown in FIG. Fig.11 In the illustrated search table, at least one target product and / or at least one target service to be recommended is stored in association with each data that can be included in the target sleep state data (each of the target property data, target rhythm data, target time data, and target sleep type data).

[0221] In this example, for the sake of simplicity, at least one target commodity and / or at least one target service to be recommended can be allocated in association with three different levels represented by each of the target property data, the target rhythm data, the target time data, and the target sleep type data. Taking the target property data as an example, at least one target commodity and / or at least one target service to be recommended can be allocated in association with level V1 (level of good sleep property of the target user), level V2 (level of standard sleep property of the target user), and level V3 (level of poor sleep property of the target user) represented by the target property data.

[0222] Levels W1 to W3 may be data indicating that the sleep rhythm of the target user is "good", "standard", and "poor" in this order. Levels X1 to X3 may be data indicating that the sleep time of the target user is "long", "standard", and "short" in this order.

[0223] Levels Y1 to Y3 may also be data indicating the sleep type of the target user (sleep) as "good", "standard", or "poor" in this order. Fig. 9 The illustrated level corresponds to “sleep type 21”, level Y2 corresponds to “sleep type 20” or “sleep type 11” in the figure, and level Y3 corresponds to “sleep type 10” in the figure.

[0224] Fig.11 The illustrated products P1 to P11 are all arbitrary products suitable for improving the user's sleep state. Fig.11 The illustrated services S1 to S8 all represent any services suitable for improving the user's sleep state. These products may include but are not limited to food, daily necessities, medicines, sports equipment, electrical products, etc. In addition, these services may include but are not limited to hospitals, clinics, gyms, etc.

[0225] For example, when the object property data in the target sleep state data indicates level V3, the server device 10A can determine product P4 and service S1 as the target product and target service, respectively. In addition, when the target sleep state data includes object time data (indicating level X2) and object rhythm data (indicating level W3), the server device 10A can determine product P8 as the target product and services S1, S3, and S4 as the target services.

[0226] Furthermore, for each data included in the target sleep state data, the total number of levels to be used, the total number of commodities and / or services to be allocated to each level, and the like can be arbitrarily determined.

[0227] In another example, the server device 10A can also use another inference model generated by performing supervised learning (machine learning) to determine at least one target product and / or at least one target service. The other inference model can be generated (updated) by performing supervised learning using data representing each level of each data included in the target sleep state data as an explanatory variable and data representing at least one target product and / or at least one target service as a target variable.

[0228] The machine learning used to learn the other inference model, for example, as described in the above-mentioned item "2A", can include but is not limited to any known machine learning, such as gradient boosting decision trees (such as LightGBM), neural networks or linear regression.

[0229] The server device 10A can output data (“recommendation data”) for identifying at least one target product and / or at least one target service from the other inference model by inputting the target sleep state data to the other inference model.

[0230] As a method of using this other estimation model, any of the following methods can be used.

[0231] The server device 10A acquires and holds the generated another estimation model, inputs the target sleep state data to the another estimation model, and acquires the advice data output from the another estimation model.

[0232] The server device 10A transmits the target sleep state data to an external device (another arbitrary server device 10 or an arbitrary terminal device 20) that acquires and stores the generated another inference model via the communication network 2, thereby causing the another inference model that receives the target sleep state data as an explanatory variable to output the recommendation data. Thereafter, the server device 10A receives the recommendation data output by the another inference model from the external device via the communication network 2.

[0233] The acquisition of such recommendation data can be performed by other server devices 10 (and / or terminal devices 20) managed by other enterprises, instead of or together with the server device 10A (and / or terminal device 20) managed by the enterprise operating the inference service, etc.

[0234] (6) Step 1010

[0235] When referring again Figure 3 Then, the server device 10A (and / or the terminal device 20) as an information processing device, for example, managed by a company operating an inference service, can provide (send) the recommendation data acquired in ST1008 to at least one terminal device 20 (or other server devices 10) managed by a company user and / or an individual user via the communication network 2. The "individual user" mentioned here can include the "target user".

[0236] Regarding the provision of recommendation data, it can be provided to at least one terminal device 20 (or other server device 10) managed by corporate users and / or individual users through any means including web pages, e-mails, chats, electronic files, recording media (USB memory, CD-ROM, etc.) and / or paper media.

[0237] Such provision of recommendation data can be performed by other server devices 10 (and / or terminal devices 20) managed by other enterprises, instead of server device 10A (and / or terminal device 20) managed by an enterprise operating an inference service, etc., or together with server device 10A (and / or terminal device 20) managed by an enterprise operating an inference service, etc.

[0238] 5. Modifications

[0239] In the above-mentioned various examples, the case where the inference model is generated and used by using the sample property data, sample rhythm data, sample time data, and sample sleep type data as target variables is described. In another example, at least one (up to three) of the sample property data, sample rhythm data, sample time data, and sample sleep type data can be used as explanatory variables instead of the target variable to generate and use the inference model.

[0240] In addition, in the above examples, Figure 3 The various steps (e.g., ST1000 to ST1010) are described as being mainly executed by a server device 10A (and / or a terminal device 20) managed by a company operating an inference service. However, any number of server devices 10 and / or a number of terminal devices 20 may be configured to share the execution of the steps. Figure 3 Various steps illustrated. Here, in one example, any of the multiple server devices 10 and / or the multiple terminal devices 20 can include a server device 10A (and / or the terminal device 20) managed by a company that operates an inference service, etc., and in another example, any of the multiple server devices 10 and / or the multiple terminal devices 20 do not include a server device 10A (and / or the terminal device 20) managed by a company that operates an inference service, etc.

[0241] In addition, in the above-mentioned "4 (4) step 1006", the definitions of the subject property data, the subject rhythm data, the subject time data, and the subject sleep type data that can be included in the subject sleep state data are shown. In another embodiment, the subject sleep state data can include at least one of the following data instead of the data shown in the above-mentioned "4 (4) step 1006".

[0242] Data used to "uniquely" identify the subject user's sleep properties ("subject property data")

[0243] Data for “uniquely” identifying the sleep rhythm of the target user (“target rhythm data”)

[0244] Data for “uniquely” identifying the sleep time of the target user (“target time data”)

[0245] Data for “uniquely” identifying the sleep type to which the target user (sleep) belongs (“target sleep type data”)

[0246] Here, "object property data" is used for "unique identification" which means that the sleep property of the object user can be identified even without relying on other data other than the "object property data". That is, in the case where multiple data collaboratively identify the sleep property of the object user, these multiple data do not meet the data used to "uniquely identify" the sleep property of the object user. For example, consider the following case: the first data contains data indicating that the sleep quality of the object user is "low quality" and a flag "0", the second data contains data indicating that the sleep quality of the object user is "standard" and a flag "0", and the third data contains data indicating that the sleep quality of the object user is "high quality" and a flag "1". In this case, the sleep property of the object user cannot be determined only by the first data and the second data. Only by checking the third data with the flag "1" can it be determined that the sleep quality of the object user is "high quality". Therefore, none of these data is data used to "uniquely" identify the sleep quality of the object user.

[0247] The same can be considered with respect to the term “uniquely identifying” included in the above definitions of “target rhythm data”, “target time data”, and “target sleep type data”.

[0248] As described above, according to the technology disclosed in the present application, by inputting the target behavior data for identifying the history of at least one behavior performed by the target user into the inference model, the target sleep state data for identifying the current (predicted) sleep state of the target user can be obtained from the inference model. In addition, at least one target product and / or at least one target service suitable for the target sleep state data thus obtained from a plurality of target products and / or a plurality of target services prepared in advance can be recommended to any user including the target user to improve the sleep state of the target user.

[0249] Thus, a computer program, an information processing device, and a method for inferring the sleep state of a target user based on a history of at least one action performed by the target user and having at least partially improved performance can be provided.

[0250] 6. Actions performed by the communication system 2 to provide menopausal status data related to the status of menopausal symptoms do

[0251] Next, refer to Fig.12 A specific example of the operation performed by the communication system 2 to provide menopausal status data related to the status of menopausal symptoms will be further described. Fig.12 It is shown by Figure 1 A flowchart of another example of the operation performed by the communication system 1 shown.

[0252] (1) Step 1100

[0253] First, in ST1100 , a server device 10A as an information processing device managed by, for example, a company operating an inference service or the like can acquire and store a plurality of sets of teaching data for generating an inference model.

[0254] Each set of teaching data constituting the plurality of sets of teaching data may include data for identifying a history of behaviors performed by a sample user corresponding to the set ("sample behavior data") and data for identifying a state of menopausal symptoms of a sample user corresponding to the set ("sample menopausal state data"). For example, the first set of teaching data may include sample behavior data for identifying a history of behaviors performed by a sample user (V) corresponding to the first set and sample menopausal state data for identifying a state of menopausal symptoms of V, and the second set of teaching data may include sample behavior data for identifying a history of behaviors performed by a sample user (W) corresponding to the second set and sample menopausal state data for identifying a state of menopausal symptoms of W.

[0255] (1A) Sample behavior data

[0256] Each sample behavior data may be data for identifying at least one (one or more) of a plurality of predetermined behaviors. Here, each of the plurality of predetermined behaviors may be a behavior that may at least partially exert a favorable or unfavorable influence on the state of menopausal symptoms of a person who performs the behavior. Such a plurality of predetermined behaviors may include, but are not limited to, the behaviors exemplified below.

[0257] (Examples of behaviors that may at least partially exert a favorable influence on a person's menopausal symptom status)

[0258] Performed aerobic exercise for at least 1 hour at a moderate pace 2 or more days in a week.

[0259] Get up at a fixed time every day.

[0260] Make sure to take some time to relax each week.

[0261] · Purchased more than 10 cans of canned beer (this behavior can be identified, for example, by data including the product category to which the beer belongs and the quantity purchased).

[0262] (Examples of behaviors that may at least partially adversely affect a person's menopausal symptom status)

[0263] Performed aerobic physical activity for at least 1 hour at a pace no more than 1 day per week.

[0264] Wake up at a different time each day.

[0265] ·Work without a single day off in a week.

[0266] · Purchased 5 or more bottles of green juice (this behavior can be identified, for example, by data including the product category to which the green juice belongs and the quantity purchased).

[0267] Each sample behavior data can be data for identifying at least one behavior among a plurality of predetermined behaviors exemplified in this manner.

[0268] In addition, in one example, the goods purchased by the user and / or the services used by the user may have a tendency to exert a favorable or unfavorable influence on the state of the user's menopausal symptoms, so the above-mentioned predetermined multiple behaviors can also include the goods purchased by the sample user (or the category to which the goods belong) and the services used by the sample user (or the category to which the services belong). Such behaviors can be generated, for example, using POS (Point Of Sales) data owned by retail stores (drug stores, supermarkets, convenience stores, etc.).

[0269] In one example, a plurality of predetermined behaviors are each assigned unique identification data (such as English letters, numbers, or a combination thereof), and each sample behavior data may include such identification data. For example, when using POS data, each of the plurality of sample behavior data may include, but is not limited to, the JICFS classification (classification code) corresponding to the purchased commodity, the commodity category, the purchase amount, the purchase quantity, the purchase frequency, and / or the purchase date, etc.

[0270] The server device 10A as an information processing device managed by, for example, a company operating an inference service or the like can acquire such sample behavior data by, for example, at least one of the following exemplified methods.

[0271] The information processing device receives POS data via the communication network 2 from at least one other server device 10 (for example, another server device 10 installed in a drug store or the like for storing POS data, or another server device 10 capable of accessing the POS data).

[0272] The information processing device receives an email or a web page recording answers to a predetermined questionnaire (questionnaire related to behavior performed by the sample user) from at least one terminal device 20 and / or at least one server device 10 via the communication network 2 .

[0273] The information processing device receives the content of the answers to the above-mentioned predetermined questionnaire from at least one server device 10 that has collected the answers from a plurality of users via the communication network 2 .

[0274] The information processing device receives the contents of the answers to the above-mentioned predetermined questionnaire from a plurality of users via a recording medium (USB memory, DVD-ROM, etc.) on which the answers are collected.

[0275] (1B) Sample menopausal status data

[0276] The sample menopausal status data included in each set of teaching data may include, but is not limited to, at least one of the data exemplified below.

[0277] Data for identifying the comprehensive status of a sample user corresponding to the group ("sample comprehensive status data")

[0278] Data for identifying the state of the first symptom of a sample user corresponding to the group ("sample first state data")

[0279] Data for identifying the state of the second symptom of a sample user corresponding to the group ("sample second state data")

[0280] First, the sample comprehensive status data can be data for identifying whether the status of the menopausal symptoms of the sample user is good (e.g., to what extent it is good) or bad (e.g., to what extent it is bad). In one example, the sample comprehensive status data can be data for identifying a score obtained by comprehensively evaluating the status of the menopausal symptoms of the sample user. In this case, the status of the menopausal symptoms of the sample user can be data for identifying a score calculated using any of the methods exemplified below.

[0281] Based on the sample users' simplified menopause index (SMI) (ref. Fig.13 ) using the method described in the SMI, or a modified score obtained by performing an operation (addition, subtraction, multiplication, division, etc.) on the score using any method.

[0282] · A score calculated based on the sample user's answers to questions in the Menopause Rating Scale using the method described in the Menopause Rating Scale, or a modified score obtained by performing calculations (addition, subtraction, multiplication, division, etc.) on the score using any method

[0283] A score calculated based on the sample user's answers to questions listed in the Kupperman Index using the method listed in the Kupperman Index, or a modified score obtained by performing calculations (addition, subtraction, multiplication, division, etc.) on the score using any method

[0284] · A score calculated based on the sample user's answers to questions in the Green Climacteric Scale using the method described in the Green Climacteric Scale, or a modified score obtained by performing calculations (addition, subtraction, multiplication, division, etc.) on the score using any method

[0285] · A score calculated based on the sample user's answers to questions in the PSST (Premenstrual Symptoms Screening Tool) using the method described in the PSST, or a modified score obtained by performing calculations (addition, subtraction, multiplication, division, etc.) on the score using any method

[0286] · A score calculated based on the sample user's answers to questions in the WHQ (Women's Health Questionnaire) using the method described in the WHQ, or a modified score obtained by performing calculations (addition, subtraction, multiplication, division, etc.) on the score using any method

[0287] · A score calculated based on the sample user's answers to questions in the VAS (Visual Analogue Scale) using the method in the VAS, or a modified score obtained by performing calculations (such as addition, subtraction, multiplication, and division) on the score using any method

[0288] · A score calculated based on the sample user's answers to questions in the HFRDI (Hot Flash Related Daily Interference Scale) using the method described in the HFRDI, or a modified score obtained by performing calculations (addition, subtraction, multiplication, division, etc.) on the score using any method

[0289] A score calculated using the method described in HFCS (Hot Flash Composite Score) based on the sample user's answers to the questions described in HFCS, or a modified score obtained by performing calculations (addition, subtraction, multiplication, division, etc.) on the score using any method

[0290] · A score calculated based on the sample user's answers to questions in MENQOL (Menopause-Specific Quality Of Life) using the method described in MENQOL, or a modified score obtained by performing calculations (addition, subtraction, multiplication, division, etc.) on the score using any method

[0291] · A score calculated using the method described in the Japanese Women's Menopausal Symptom Evaluation Table based on the sample users' answers to the questions described in the Japanese Women's Menopausal Symptom Evaluation Table, or a modified score obtained by performing calculations (such as addition, subtraction, multiplication, and division) on the score using any method

[0292] A score calculated based on the sample user's answers to questions described in another arbitrary questionnaire or scale using a method described in the questionnaire or scale, or a corrected score obtained by calculating the score using an arbitrary method.

[0293] Second, the sample first status data can be data for identifying whether the status of a symptom of the sample user among multiple menopausal symptoms is good (e.g., to what extent it is good) or bad (e.g., to what extent it is bad). In one example, the sample first status data can be data for identifying a score obtained by evaluating the status of the above-mentioned symptom of the sample user.

[0294] If SMI is taken as an example, the above-mentioned multiple menopausal symptoms can be multiple menopausal symptoms recorded in SMI, such as "hot face", "easy sweating", "cold waist, hands and feet". In addition, the above-mentioned multiple menopausal symptoms can also be multiple menopausal symptoms recorded in any index listed as above, in addition to the symptoms recorded in SMI. In addition, the above-mentioned one symptom can be any one symptom selected from such multiple menopausal symptoms.

[0295] If SMI is taken as an example, in one example, the sample first state data can be obtained from the sample user for SMI (refer to Fig.13 ) is extracted from the answers to the questions recorded in ). In this case, the sample first state data is, for example, data for identifying the state of a symptom such as "shortness of breath, palpitations", for example, it can be used to identify a score of 0, 4, 8 or 12 (refer to Fig.13 In another example, the sample first state data can be obtained from the sample user via email or web page.

[0296] Third, the sample second status data can be data for identifying whether the status of another symptom of the sample user among the multiple menopausal symptoms is good (e.g., to what extent it is good) or bad (e.g., to what extent it is bad). In one example, the sample second status data can be data for identifying a score obtained by evaluating the status of the above-mentioned another symptom of the sample user.

[0297] If SMI is taken as an example, the above-mentioned multiple menopausal symptoms can be multiple menopausal symptoms recorded in SMI, such as "hot face", "easy sweating", "cold waist, hands and feet". In addition, the above-mentioned multiple menopausal symptoms can also be multiple menopausal symptoms recorded in any index listed as above, in addition to the symptoms recorded in SMI. In addition, the above-mentioned another symptom can be any one symptom selected from such multiple menopausal symptoms, but can be a symptom different from any of the above-mentioned symptoms.

[0298] If SMI is taken as an example, in one example, the sample second state data can be obtained from the sample user for SMI (refer to Fig.13 ) is extracted from the answers to the questions recorded in ). In this case, the sample second state data is, for example, data for identifying the state of another symptom such as "easy fatigue", for example, it can be used to identify a score of 0, 2, 4 or 7 (refer to Fig.13 In another example, the sample second state data can be obtained from the sample user via email or web page.

[0299] The server device 10A (and / or the terminal device 20) as an information processing device, for example, managed by a company operating an inference service, can combine such sample behavior data with sample menopausal status data for each sample user, and generate and store a set of teaching data for the sample user. To achieve this, for example, the information processing device can extract sample behavior data with identification data for identifying a sample user from the acquired multiple sample behavior data, extract sample menopausal status data with identification data for identifying a sample user who is the same user as the sample user from the acquired multiple sample menopausal status data, and combine them to generate and save a set of teaching data related to the sample user. By performing the same processing on multiple sample users, the information processing device can generate and save multiple sets of teaching data.

[0300] (2) Step 1102

[0301] Return to Figure 3 Then, in ST1102, the server device 10A (and / or the terminal device 20) as an information processing device, such as managed by a company operating an inference service, uses the multiple sets of teaching data acquired in ST1100 to make the learning model learn through machine learning, thereby generating an inference model. The machine learning mentioned here can include but is not limited to any known machine learning, such as a gradient boosting decision tree (such as LightGBM), a neural network, or a linear regression.

[0302] (2A) The first plurality of sets of teaching data included in the plurality of sets of teaching data

[0303] In one example, the information processing device can generate (update) an inference model using the first plurality of teaching data included in the plurality of teaching data. Each set of teaching data included in the first plurality of teaching data can include sample behavior data of a sample user corresponding to the set and sample menopausal status data of a sample user corresponding to the set. Here, the sample menopausal status data in each set of teaching data can include sample comprehensive status data of a sample user corresponding to the set.

[0304] In this case, the information processing device can input the first plurality of groups of teaching data (each group of teaching data includes sample behavior data of a sample user corresponding to the group and sample comprehensive state data of the sample user) into the learning model to generate (update) an inference model. Such an inference model can output at least object comprehensive state data for identifying the comprehensive state of the object user in response to object behavior data for identifying the history of behaviors performed by the object user as an explanatory variable as described later.

[0305] Furthermore, in one embodiment, the following case is considered: the sample menopausal status data in each set of teaching data is data for identifying a “corresponding score” of the comprehensive status of menopausal symptoms of a sample user corresponding to the set.

[0306] In this case, the information processing device can perform preprocessing on each group of teaching data included in the first plurality of groups of teaching data. Specifically, the information processing device can perform the following preprocessing on each group of teaching data: using at least one threshold value, the score identified by the sample comprehensive state data corresponding to the group is transformed into one of a plurality of categories. For example, when the information processing device uses one threshold value (first threshold value), the score identified by the sample comprehensive state data corresponding to each group can be transformed into one of two categories ("high" and "low") according to whether the score is above the first threshold value or below the first threshold value. Or, for example, when the information processing device uses two threshold values ​​(a first threshold value and a second threshold value that is smaller than the first threshold value), the score identified by the sample comprehensive state data corresponding to each group can also be transformed into one of three categories ("high", "medium" and "low") according to whether the score is above the first threshold value, above the second threshold value and below the first threshold value, or below the second threshold value.

[0307] In this way, the information processing device can perform the following preprocessing on each group of teaching data contained in the first plurality of groups of teaching data: transforming the sample comprehensive status data of the "corresponding score" of the comprehensive status of menopausal symptoms of a sample user corresponding to the group into the sample comprehensive status data of the "corresponding category" of the comprehensive status of menopausal symptoms of a sample user corresponding to the group.

[0308] Thus, the information processing device can input the preprocessed first multiple groups of teaching data (each group of teaching data contains sample behavior data of the sample user corresponding to the group and sample comprehensive status data of the sample user) into the learning model to generate (update) an inference model.

[0309] Such an inference model can output at least object comprehensive state data for identifying a category corresponding to the comprehensive state of menopausal symptoms of the object user in response to object behavior data for identifying a history of behaviors performed by the object user being input as an explanatory variable as described later.

[0310] (2B) The second plurality of sets of teaching data included in the plurality of sets of teaching data

[0311] In one example, the information processing device can generate (update) an inference model using a second plurality of sets of teaching data included in the plurality of sets of teaching data. Each set of teaching data included in the second plurality of sets of teaching data can include sample behavior data of a sample user corresponding to the set and sample menopausal status data of a sample user corresponding to the set. Here, the sample menopausal status data in each set of teaching data can include sample first status data of a sample user corresponding to the set.

[0312] In this case, the information processing device can input a second plurality of groups of teaching data (each group of teaching data includes sample behavior data of a sample user corresponding to the group and sample first state data of the sample user) into the learning model to generate (update) an inference model. Such an inference model can output at least object first state data for identifying the state of the first symptom of the object user in response to object behavior data of a history record of a behavior performed by the object user being input as an explanatory variable as described later.

[0313] Furthermore, in one embodiment, the following case is considered: the sample first state data in each set of teaching data is data for identifying a “corresponding score” of the state of the first symptom of a sample user corresponding to the set.

[0314] In this case, the information processing device can perform preprocessing on each group of teaching data included in the second plurality of groups of teaching data. Specifically, the information processing device can perform the following preprocessing on each group of teaching data: using at least one threshold value, the score identified by the sample first state data corresponding to the group is transformed into one of a plurality of categories. For example, when the information processing device uses one threshold value (first threshold value), the score identified by the sample first state data corresponding to each group can be transformed into one of two categories ("high" and "low") according to whether the score is above the first threshold value or below the first threshold value. Or, for example, when the information processing device uses two threshold values ​​(a first threshold value and a second threshold value smaller than the first threshold value), the score identified by the sample first state data corresponding to each group can be transformed into one of three categories ("high", "medium" and "low") according to whether the score is above the first threshold value, above the second threshold value and below the first threshold value, or below the second threshold value.

[0315] In this way, the information processing device can perform the following preprocessing on each group of teaching data contained in the second plurality of groups of teaching data: transforming the sample first state data used to identify the "corresponding score" of the state of the first symptom of a sample user corresponding to the group into the sample first state data used to identify the "corresponding category" of the state of the first symptom of a sample user corresponding to the group.

[0316] Thus, the information processing device can input the preprocessed second plurality of groups of teaching data (each group of teaching data includes sample behavior data of the sample user corresponding to the group and sample first state data of the sample user) into the learning model to generate (update) an inference model.

[0317] Such an inference model can output at least object first state data for identifying a category corresponding to the state of the first symptom of the object user in response to object behavior data for identifying a history of behaviors performed by the object user being input as an explanatory variable as described later.

[0318] Furthermore, in addition to the preprocessing described above (recorded as "preprocessing A" for convenience), the information processing device can also perform further preprocessing (recorded as "preprocessing B" for convenience). Specifically, first, the information processing device can perform "preprocessing B" such as extracting other sets of teaching data from the second plurality of sets of teaching data. Each set of teaching data included in the other plurality of sets of teaching data has sample first state data for identifying scores above a threshold.

[0319] For example, if the first symptom is SMI (refer to Fig.13), each set of teaching data included in the other plurality of sets of teaching data may be sample first state data for identifying a score (10 or 6) above the threshold value of 5. That is, each set of teaching data included in the other plurality of sets of teaching data includes teaching data only related to sample users with a stronger first symptom (here, "easy sweating").

[0320] Furthermore, the information processing device can set the other multiple sets of teaching data extracted in this way as the "new" second multiple sets of teaching data. After that, the information processing device can perform the above-mentioned preprocessing A on the "new" second multiple sets of teaching data. In the preprocessing A, compared with the above-mentioned example without performing preprocessing B, at least one threshold value to be used can be set to a larger value. Next, the information processing device can input the "new" second multiple sets of teaching data after preprocessing A into the learning model to generate (update) the inference model.

[0321] This "new" second plurality of sets of teaching data includes sample behavior data and sample first state data related only to sample users with stronger first symptoms. Therefore, the inference model learned using this "new" second plurality of sets of teaching data can more easily infer that a target user who has performed a behavior that is likely to be associated with a strong first symptom is a user with a strong first symptom, and can also infer the level of intensity of the first symptom of such a target user in detail.

[0322] (2C) The third plurality of teaching data included in the plurality of teaching data

[0323] In one example, the information processing device can generate (update) an inference model using a third plurality of sets of teaching data included in the plurality of sets of teaching data. Each set of teaching data included in the third plurality of sets of teaching data can include sample behavior data of a sample user corresponding to the set and sample menopausal status data of a sample user corresponding to the set. Here, the sample menopausal status data in each set of teaching data can include sample second status data of a sample user corresponding to the set.

[0324] In this case, the information processing device can input a third plurality of groups of teaching data (each group of teaching data includes sample behavior data of a sample user corresponding to the group and sample second state data of the sample user) into the learning model to generate (update) an inference model. Such an inference model can output at least object second state data for identifying the state of the second symptom of the object user in response to object behavior data of a history record of a behavior performed by the object user being input as an explanatory variable as described later.

[0325] Furthermore, in one embodiment, the following case is considered: the sample second state data in each set of teaching data is data for identifying “a corresponding score” of the state of the second symptom of a sample user corresponding to the set.

[0326] In this case, the information processing device can perform preprocessing on each group of teaching data included in the third plurality of groups of teaching data. Specifically, the information processing device can perform the following preprocessing on each group of teaching data: using at least one threshold value, the score identified by the sample second state data corresponding to the group is transformed into one of a plurality of categories. For example, when the information processing device uses one threshold value (first threshold value), the score identified by the sample second state data corresponding to each group can be transformed into one of two categories ("high" and "low") according to whether the score is above the first threshold value or below the first threshold value. Or, for example, when the information processing device uses two threshold values ​​(a first threshold value and a second threshold value smaller than the first threshold value), the score identified by the sample second state data corresponding to each group can be transformed into one of three categories ("high", "medium" and "low") according to whether the score is above the first threshold value, above the second threshold value and below the first threshold value, or below the second threshold value.

[0327] In this way, the information processing device can perform the following preprocessing on each group of teaching data contained in the third plurality of groups of teaching data: transforming the sample second state data used to identify the "corresponding score" of the state of the second symptom of a sample user corresponding to the group into the sample second state data used to identify the "corresponding category" of the state of the second symptom of a sample user corresponding to the group.

[0328] Thus, the information processing device can input the preprocessed third multiple groups of teaching data (each group of teaching data contains sample behavior data of the sample user corresponding to the group and sample first state data of the sample user) into the learning model to generate (update) the inference model.

[0329] Such an inference model can output at least object second state data for identifying a category corresponding to the state of the second symptom of the object user in response to object behavior data for identifying a history of behaviors performed by the object user being input as an explanatory variable as described later.

[0330] Furthermore, in addition to the preprocessing described above (recorded as "preprocessing A" for convenience), the information processing device can also perform further preprocessing (recorded as "preprocessing B" for convenience). Specifically, first, the information processing device can perform "preprocessing B" such as extracting other sets of teaching data from the third set of teaching data. Each set of teaching data included in the other set of teaching data has sample second state data for identifying scores above a threshold.

[0331] For example, in cases where the second symptom is SMI (refer to Fig.13 ), each set of teaching data included in the other plurality of sets of teaching data can be sample second state data for identifying a score (7 or 4) above the threshold value of 3. That is, each set of teaching data included in the other plurality of sets of teaching data includes teaching data only related to sample users having a stronger second symptom (here, "easy to get tired").

[0332] Furthermore, the information processing device can set the other multiple sets of teaching data extracted in this way as "new" third multiple sets of teaching data. After that, the information processing device can perform the above-mentioned preprocessing A on the "new" third multiple sets of teaching data. In the preprocessing A, compared with the above-mentioned example without performing preprocessing B, at least one threshold value to be used can be set to a larger value. Next, the information processing device can input the "new" third multiple sets of teaching data after preprocessing A into the learning model to generate (update) the inference model.

[0333] This "new" third plurality of sets of teaching data includes sample behavior data and sample second state data related only to sample users with stronger second symptoms. Therefore, the inference model learned using this "new" third plurality of sets of teaching data can more easily infer that a target user who has performed a behavior that is likely to be associated with a stronger second symptom is a user with a stronger second symptom, and can also infer the level of intensity of the second symptom of such a target user in detail.

[0334] The generation (update) of the above-described inference model can be performed by another server device 10 (and / or terminal device 20) managed by another enterprise, instead of the server device 10A (and / or terminal device 20) managed by the enterprise operating the inference service, etc., or together with the server device 10A (and / or terminal device 20) managed by the enterprise operating the inference service, etc. The inference model generated (updated) by the other server device 10 (terminal device 20) can be transmitted to any server device 10 including the server device 10A, or can be accessed and used by any server device 10 including the server device 10A.

[0335] In addition, the names such as "first" to "third" marked in front of "multiple sets of teaching data" are used to conveniently distinguish "multiple sets of teaching data" marked with one of these names from "multiple sets of teaching data" marked with other of these names, and do not limit any order or content.

[0336] (3) Step 1104

[0337] When referring again Fig.12 When, next, in ST1104, the server device 10A (and / or the terminal device 20) as an information processing device, for example, managed by a company operating an inference service, etc., can obtain target behavior data for identifying a history of behaviors performed by a target user who is the target of inference, for example, from the terminal device 20 (or other server device 10) that accesses the server device 10A by executing a specific application (an application for receiving provision of the inference service, etc.), and / or the terminal device 20 that accesses the server device 10A by executing a browser. The target behavior data can be data for identifying at least one (one or more) behaviors performed by the target user.

[0338] Such acquisition of object behavior data can be performed by other server devices 10 (and / or terminal devices 20) managed by other enterprises, instead of server device 10A (and / or terminal device 20) managed by an enterprise operating an inference service, etc., or together with server device 10A (and / or terminal device 20) managed by an enterprise operating an inference service, etc.

[0339] (4) Step 1106

[0340] Next, in ST1106, the server device 10A (and / or the terminal device 20) as an information processing device, for example, managed by a company operating an inference service, can use the object behavior data acquired in ST1104 and the inference model generated in ST1102 to acquire the object menopausal status data for identifying the state of the menopausal symptoms of the object user. Specifically, the server device 10A can output the object menopausal status data as a target variable from the inference model by inputting the object behavior data as an explanatory variable into the inference model.

[0341] The target menopausal status data may include at least one of the following data.

[0342] Data for identifying the comprehensive status of menopausal symptoms of the target user ("target comprehensive status data")

[0343] Data for identifying the state of the first symptom of the target user ("target first state data")

[0344] Data for identifying the state of the second symptom of the target user ("target second state data")

[0345] Which data the inference model outputs as the target menopausal status data may depend on which data was used as the sample menopausal status data when generating (updating) the inference model.

[0346] For example, when the first plurality of teaching data are used as sample menopausal status data when generating (updating) the inference model (item "6(2)(2A)" above), the inference model can at least output the object comprehensive status data as the object menopausal status data.

[0347] For example, when the second plurality of teaching data are used as sample menopausal status data when generating (updating) the inference model (item "6(2)(2B)" above), the inference model can output at least the object first status data as the object menopausal status data.

[0348] For example, when the third plurality of teaching data are used as sample menopausal status data when generating (updating) the inference model (the above-mentioned "6(2)(2C)" item), the inference model can output at least the object second status data as the object menopausal status data.

[0349] For example, when the first plurality of teaching data sets are used as sample menopausal state data to generate (update) the inference model (the above-mentioned "6(2)(2A)" item), and then the second plurality of teaching data sets are used as sample menopausal state data to generate (update) the inference model (the above-mentioned "6(2)(2B)" item), the inference model can output at least the object comprehensive state data and the object first state data as the object menopausal state data. And, when the third plurality of teaching data sets are used as sample menopausal state data to generate (update) the inference model (the above-mentioned "6(2)(2C)" item), the inference model can output at least the object comprehensive state data, the object first state data, and the object second state data as the object menopausal state data.

[0350] As a method of using the estimation model, any of the methods exemplified below can be used.

[0351] The server device 10A acquires and stores the inference model generated in ST1104, inputs the subject behavior data to the inference model, and acquires the subject menopausal state data output from the inference model.

[0352] The server device 10A transmits the target behavior data to an external device (another arbitrary server device 10 or an arbitrary terminal device 20) that acquires and stores the inference model generated in ST1104 via the communication network 2, thereby causing the inference model that has been input with the target behavior data as an explanatory variable to output the target menopausal status data. Thereafter, the server device 10A receives the target menopausal status data output by the inference model from the external device via the communication network 2.

[0353] Such acquisition of the menopausal status data of the object can be performed by other server devices 10 (and / or terminal devices 20) managed by other enterprises, instead of the server device 10A (and / or terminal device 20) managed by the enterprise operating the inference service, etc., or together with the server device 10A (and / or terminal device 20) managed by the enterprise operating the inference service, etc.

[0354] (5) Step 1108

[0355] Next, in ST1108, the server device 10A (and / or the terminal device 20) as an information processing device, for example, managed by a company operating an inference service, can use the target menopausal status data acquired in ST1106 to determine at least one target product / one target service corresponding to the target menopausal status data. Such a target product / target service can be suitable for the current (inferred) menopausal symptom status of the target user identified by the target menopausal status data.

[0356] Specifically, in one example, the server device 10A can first prepare and store a search table that stores the target menopausal status data in association with multiple products and / or multiple services. Next, the server device 10A can input the target menopausal status data acquired in ST1106 into the search table, and can acquire data ("recommendation data") for identifying at least one target product and / or at least one target service from the search table.

[0357] Fig.14 is conceptually shown in Figure 1 FIG. 1 is a diagram showing another example of a search table used in the communication system 1. Fig.14 In the illustrated search table, at least one target product and / or at least one target service to be recommended is stored in association with each data that can be included in the target menopausal status data (each data of the target comprehensive status data, the target first status data, and the target second status data).

[0358] In this example, for the sake of simplicity, at least one target product and / or at least one target service to be recommended can be allocated in association with each of the three different levels represented by each of the target comprehensive status data, the target first status data, and the target second status data. If the target comprehensive status data is taken as an example, at least one target product and / or at least one target service to be recommended can be allocated in association with level V1 (level of good comprehensive status of menopausal symptoms of the target user), level V2 (level of standard comprehensive status of menopausal symptoms of the target user), and level V3 (level of poor comprehensive status of menopausal symptoms of the target user) represented by the target comprehensive status data.

[0359] The levels W1 to W3 may be data indicating that the state of the first symptom of the target user is "good", "standard", or "poor" in this order. The levels X1 to X3 may be data indicating that the state of the second symptom of the target user is "good", "standard", or "poor" in this order.

[0360] Fig.14 The exemplified products (products with reference numerals consisting of "P" and numbers such as P10 and P20) are all products suitable for improving the condition of the user's menopausal symptoms. Fig.14 The illustrated services (such as S10 and S20, which are marked with reference numerals consisting of "S" and numbers) all represent any services suitable for improving the state of the user's menopausal symptoms. These products can include but are not limited to food, daily necessities, medicines, sports equipment, electrical products, etc. In addition, these services can include but are not limited to hospitals, clinics, gyms, etc.

[0361] For example, when the target comprehensive status data in the target menopausal status data indicates level V3, the server device 10A can determine product P40 and service S10 as the target product and target service, respectively. In addition, when the target menopausal status data includes target second status data (indicating level X2) and target first status data (indicating level W3), the server device 10A can determine product P80 as the target product and services S10, S30, and S40 as the target services.

[0362] Furthermore, for each data included in the target menopausal status data, the total number of levels to be used, the total number of commodities and / or services to be allocated to each level, and the like can be arbitrarily determined.

[0363] In another example, the server device 10A can also use another inference model generated by performing supervised learning (machine learning) to determine at least one target product and / or at least one target service. The other inference model is generated (updated) by performing supervised learning using data representing each level of each data included in the target menopausal status data as explanatory variables and data representing at least one target product and / or at least one target service as target variables.

[0364] The machine learning used to learn the other inference model, for example, as described above, can include but is not limited to any known machine learning, such as a gradient boosting decision tree (eg, LightGBM), a neural network, or linear regression.

[0365] The server device 10A can output data (“recommendation data”) for identifying at least one target commodity and / or at least one target service from the other inference model by inputting the target menopausal status data to the other inference model.

[0366] As a method of using this other estimation model, any of the following methods can be used.

[0367] The server device 10A acquires and stores the generated another estimation model, inputs the target menopausal state data to the another estimation model, and acquires the advice data output from the another estimation model.

[0368] The server device 10A transmits the target menopausal status data to an external device (another arbitrary server device 10 or an arbitrary terminal device 20) that acquires and stores the generated another inference model via the communication network 2, thereby causing the another inference model that has the target menopausal status data input as an explanatory variable to output the recommendation data. Thereafter, the server device 10A receives the recommendation data output by the another inference model from the external device via the communication network 2.

[0369] Such acquisition of recommendation data can be performed by other server devices 10 (and / or terminal devices 20) managed by other enterprises, instead of server device 10A (and / or terminal device 20) managed by an enterprise operating an inference service, etc., or together with server device 10A (and / or terminal device 20) managed by an enterprise operating an inference service, etc.

[0370] (6) Step 1110

[0371] When referring again Fig.12Then, the server device 10A (and / or the terminal device 20) as an information processing device, for example, managed by a company operating an inference service, can provide (send) the recommendation data acquired in ST1108 to at least one terminal device 20 (or other server devices 10) managed by a company user and / or an individual user via the communication network 2. The "individual user" mentioned here can include the "target user".

[0372] Regarding the provision of recommendation data, it can be provided to at least one terminal device 20 (or other server device 10) managed by corporate users and / or individual users through any means including web pages, e-mails, chats, electronic files, recording media (USB memory, CD-ROM, etc.) and / or paper media.

[0373] Such provision of recommendation data can be performed by other server devices 10 (and / or terminal devices 20) managed by other enterprises, instead of server device 10A (and / or terminal device 20) managed by an enterprise operating an inference service, etc., or together with server device 10A (and / or terminal device 20) managed by an enterprise operating an inference service, etc.

[0374] 7. Modifications

[0375] In the above-mentioned various examples, the case where the sample comprehensive state data, the sample first state data, and the sample second state data are used as target variables to generate and use the inference model is described. In another example, at least one (maximum 2) of the sample comprehensive state data, the sample first state data, and the sample second state data can be used as an explanatory variable instead of a target variable to generate and use the inference model.

[0376] In addition, in the above examples, Fig.12 The various steps (e.g., ST1100 to ST1110) are described as being mainly executed by a server device 10A (and / or a terminal device 20) managed by a company operating an inference service. However, any number of server devices 10 and / or a number of terminal devices 20 may be configured to share the execution of the steps. Fig.12 Various steps illustrated. Here, in one example, any of the multiple server devices 10 and / or the multiple terminal devices 20 can include a server device 10A (and / or the terminal device 20) managed by a company that operates an inference service, etc., and in another example, any of the multiple server devices 10 and / or the multiple terminal devices 20 do not include a server device 10A (and / or the terminal device 20) managed by a company that operates an inference service, etc.

[0377] As described above, according to the technology disclosed in the present application, by inputting the object behavior data for identifying the history of at least one behavior performed by the object user into the inference model, the object menopausal status data for identifying the current (predicted) state of the menopausal symptoms of the object user can be obtained from the inference model. In addition, at least one object product and / or at least one object service suitable for the object menopausal status data obtained in this way among the plurality of object products and / or plurality of object services prepared in advance can be recommended to any user including the object user to improve the menopausal symptoms of the object user.

[0378] Thus, a computer program, an information processing device, and a method for inferring the state of menopausal symptoms of a target user based on a history of at least one behavior performed by the target user and having at least partially improved performance can be provided.

[0379] 8. About inference models for predicting various health states

[0380] The technology disclosed in the present application can be applied not only to predicting the above-mentioned sleep state and menopausal symptoms, but also to predicting various health conditions as exemplified below.

[0381] Movement status

[0382] Immune status

[0383] Hydration status

[0384] Nutritional status

[0385] (1) Regarding the state of motion

[0386] The inference model for predicting the exercise state is used to predict the state associated with the lifestyle disease, etc. In order to generate the inference model, the sample behavior data as the explanatory variable can be at least one of the plurality of data exemplified below.

[0387] Data used to identify the products (or the categories to which the products belong) purchased by sample users within a certain period (e.g., one year)

[0388] Data used to identify the services (or the categories to which the services belong) used by sample users within a certain period (e.g., one year)

[0389] Data used to identify whether the sample user is obese

[0390] Data for identifying the weight change of sample users over a certain period of time (e.g., from their 30s to the present)

[0391] Data used to identify the amount of exercise (e.g., number of steps) of sample users

[0392] In addition, the data for identifying the symptom as the target variable can be at least one of the plurality of data (which may also be a measured value) exemplified below.

[0393] Data used to identify the sample user’s weight

[0394] ·BMI (Body Mass Index) data used to identify sample users

[0395] Data on the number of steps used to identify sample users

[0396] Data used to identify whether sample users smoke

[0397] ·Data used to identify the diagnostic baseline value of metabolic syndrome for sample users

[0398] Furthermore, the metabolic syndrome diagnosis reference value includes visceral fat accumulation and / or waist circumference of the sample user, and at least two of the multiple values ​​exemplified below.

[0399] ·Thresholds specified for diagnosing hypertriglyceridemia and / or low high-density lipoprotein cholesterol in sample users

[0400] Systolic and / or diastolic blood pressure of sample users

[0401] ·Thresholds defined for diagnosing fasting hyperglycemia for sample users

[0402] The information processing device can generate an inference model for predicting the motion state by performing machine learning using such explanatory variables and target variables. The actions performed by the communication system 2 in association with the generation and utilization method of such an inference model can be substantially equivalent to the actions described in the above-mentioned items "6" and "7" in association with the method of predicting the state of the above-mentioned menopausal symptoms. In the above-mentioned items "6" and "7" (especially referring to Fig.12 and Fig.14 ) In the description described in the foregoing, by replacing the explanatory variables and target variables used in the inference model for predicting the state of menopausal symptoms with the explanatory variables and target variables described in this item "8(1)", the actions performed by the communication system 2 in association with the generation and utilization method of the inference model for predicting the motion state will be understood by those skilled in the art.

[0403] As described above, according to the technology disclosed in the present application, by inputting object behavior data for identifying the history of at least one behavior performed by the object user into the inference model, object state data for identifying the current (predicted) motion state of the object user can be obtained from the inference model. Furthermore, at least one object product and / or at least one object service suitable for the object state data obtained in this way among a plurality of object products and / or a plurality of object services prepared in advance can be recommended to any user including the object user to improve the motion state of the object user.

[0404] Thus, a computer program, an information processing device, and a method for inferring a motion state of a target user based on a history of at least one action performed by the target user and having at least partially improved performance can be provided.

[0405] (2) Regarding immune status

[0406] The inference model for predicting immune status is used to predict the degree of symptoms of a person whose body's immunity has decreased. In order to generate the inference model, the sample behavior data as an explanatory variable can be at least one of the multiple data exemplified below.

[0407] Data used to identify the products (or the categories to which the products belong) purchased by sample users within a certain period (e.g., one year)

[0408] Data used to identify the services (or the categories to which the services belong) used by sample users within a certain period (e.g., one year)

[0409] In addition, the data for identifying the symptom as the target variable can be at least one of the plurality of data (which may also be a measured value) exemplified below.

[0410] Data used to identify sample users’ susceptibility to colds

[0411] Data used to identify the frequency of colds among sample users

[0412] ・Data on the concentration of SIgA (Secretory Immunoglobulin A) used to identify the sample user

[0413] Data used to identify allergy symptoms of sample users

[0414] Data used to identify the sample user’s oral environment and / or immune status

[0415] Data used to identify the sleeping state of sample users

[0416] Data used to identify the motion status of sample users

[0417] Data used to identify the stress status of sample users

[0418] In addition, the data used to identify such symptoms can be either the subjective symptoms of sample users obtained from a questionnaire survey, etc., or data obtained from the reception history of sample users in a medical institution. The data used to identify the symptom can be data for identifying one of multiple (preferably more than three) categories.

[0419] The information processing device can generate an inference model for predicting the immune status by performing machine learning using such explanatory variables and target variables. The actions performed by the communication system 2 in association with the generation and utilization methods of such inference models can be substantially equivalent to the actions described in the above-mentioned items "6" and "7" in association with the method of predicting the status of the above-mentioned menopausal symptoms. In the above-mentioned items "6" and "7" (especially with reference to Fig.12 and Fig.14 ) described above, by replacing the explanatory variables and target variables used in the inference model for predicting the state of menopausal symptoms with the explanatory variables and target variables described in this item "8(2)", the actions performed by the communication system 2 in association with the generation and utilization method of the inference model for predicting the immune status will be understood by those skilled in the art.

[0420] As described above, according to the technology disclosed in the present application, by inputting object behavior data for identifying the historical records of at least one behavior performed by the object user into the inference model, object status data for identifying the current (predicted) immune status of the object user can be obtained from the inference model. Furthermore, at least one object product and / or at least one object service suitable for the object status data obtained in this way among a plurality of object products and / or a plurality of object services prepared in advance can be recommended to any user including the object user to improve the immune status of the object user.

[0421] Thus, a computer program, an information processing device, and a method for inferring the immune status of a target user based on a history of at least one behavior performed by the target user and having at least partially improved performance can be provided.

[0422] (3) Regarding hydration status

[0423] The inference model for predicting the hydration state is used to predict the degree of symptoms of a person whose body immunity is reduced. In order to generate the inference model, the sample behavior data as the explanatory variable can be at least one of the multiple data exemplified below.

[0424] Data used to identify the products (or the categories to which the products belong) purchased by sample users within a certain period (e.g., one year)

[0425] Data used to identify the services (or the categories to which the services belong) used by sample users within a certain period (e.g., one year)

[0426] In addition, the data for identifying the symptom as the target variable can be at least one of the plurality of data (which may also be a measured value) exemplified below.

[0427] Data on serum sodium values ​​used to identify sample users

[0428] ·Data on urine osmolality used to identify the sample user

[0429] · Urine specific gravity data used to identify the sample user

[0430] Data used to identify the sample user’s urine color

[0431] Data on BUN (urea nitrogen) / creatinine ratio used to identify sample users

[0432] Data on the Dehydration Rating Scale used to identify sample users

[0433] Data for identifying subjective symptoms associated with dehydration among sample users

[0434] These data can be generated based on clinical examination values ​​received by the sample user through health diagnosis or physical examination, or based on the sample user's answers to a health questionnaire. In addition, data for identifying subjective symptoms caused by insufficient water in the sample user can be generated based on the sample user's answers to the questionnaire. The degree of the subjective symptoms can be grasped based on serum sodium value, urine osmotic pressure, urine specific gravity, urine color, BUN / creatinine ratio, water intake and / or physical activity, etc.

[0435] The information processing device can generate an inference model for predicting the water replenishment state by performing machine learning using such explanatory variables and target variables. The actions performed by the communication system 2 in association with the generation and utilization method of such an inference model can be substantially equivalent to the actions described in the above-mentioned items "6" and "7" in association with the method of predicting the state of the above-mentioned menopausal symptoms. In the above-mentioned items "6" and "7" (especially referring to Fig.12 and Fig.14) described above, by replacing the explanatory variables and target variables used in the inference model for predicting the state of menopausal symptoms with the explanatory variables and target variables described in this item "8(3)", the actions performed by the communication system 2 in association with the generation and utilization method of the inference model for predicting the state of hydration replenishment will be understood by those skilled in the art.

[0436] As described above, according to the technology disclosed in the present application, by inputting the object behavior data for identifying the history of at least one behavior performed by the target user into the inference model, the object status data for identifying the current (predicted) hydration status of the target user can be obtained from the inference model. Furthermore, at least one target product and / or at least one target service suitable for the object status data thus obtained from a plurality of target products and / or a plurality of target services prepared in advance can be recommended to any user including the target user, so as to improve the hydration status of the target user.

[0437] Thus, a computer program, an information processing device, and a method for inferring the hydration status of a target user based on a history of at least one action performed by the target user and having at least partially improved performance can be provided.

[0438] (4) Nutritional status

[0439] The inference model for predicting nutritional status is used to predict the degree of symptoms of a person who fails to replenish nutrition. In order to generate the inference model, the sample behavior data as an explanatory variable can be at least one of the multiple data exemplified below.

[0440] Data used to identify the products (or the categories to which the products belong) purchased by sample users within a certain period (e.g., one year)

[0441] Data used to identify the services (or the categories to which the services belong) used by sample users within a certain period (e.g., one year)

[0442] In addition, the data for identifying the symptom as the target variable can be at least one of the plurality of data (which may also be a measured value) exemplified below.

[0443] Data for identifying the presence or absence of subjective symptoms associated with decreased nutrition among sample users

[0444] Data used to identify the Diversity Score (DVS) of the food intake of sample users

[0445] Data on the Simplified Nutritional Appetite Questionnaire (SNAQ) used to identify sample users

[0446] BMI (Body Mass Index) data used to identify sample users

[0447] In addition, the data used to identify such symptoms can be either the subjective symptoms of sample users obtained from questionnaires, etc., or data obtained from the reception history of sample users in medical institutions. The data used to identify the symptom can be data for identifying one of multiple (preferably more than three) categories.

[0448] The information processing device can generate an inference model for predicting nutritional status by performing machine learning using such explanatory variables and target variables. The actions performed by the communication system 2 in association with the generation and utilization methods of such inference models can be substantially equivalent to the actions described in the above-mentioned items "6" and "7" in association with the method of predicting the state of the above-mentioned menopausal symptoms. In the above-mentioned items "6" and "7" (especially with reference to Fig.12 and Fig.14 ) In the description described in the foregoing, by replacing the explanatory variables and target variables used in the inference model for predicting the state of menopausal symptoms with the explanatory variables and target variables described in this item "8(4)", the actions performed by the communication system 2 in association with the generation and utilization method of the inference model for predicting the nutritional status will be understood by those skilled in the art.

[0449] As described above, according to the technology disclosed in the present application, by inputting the subject behavior data for identifying the history of at least one behavior performed by the subject user into the inference model, the subject status data for identifying the current (predicted) nutritional status of the subject user can be obtained from the inference model. Furthermore, at least one subject product and / or at least one subject service suitable for the subject status data thus obtained from a plurality of pre-prepared subject products and / or a plurality of subject services can be recommended to any user including the subject user to improve the nutritional status of the subject user.

[0450] Thus, a computer program, an information processing apparatus, and a method for inferring the nutritional status of a subject user based on a history of at least one behavior performed by the subject user and having at least partially improved performance can be provided.

[0451] 9. Various methods

[0452] The computer program involved in the first embodiment is capable of "being executed by at least one processor to cause the at least one processor to perform the following functions: obtaining object behavior data for identifying a history of behaviors performed by a target user, and outputting object sleep state data for identifying a sleep state of the target user from the inference model by inputting the object behavior data into an inference model generated by performing supervised learning, the object sleep state data including at least one of object property data for identifying a sleep property of the target user, object rhythm data for identifying a sleep rhythm of the target user, object time data for identifying a sleep time of the target user, and object type data for identifying a sleep type to which the target user belongs."

[0453] The computer program involved in the second method is capable of "enabling the at least one processor to perform the following function in the above-mentioned first method: obtaining the inference model generated by inputting multiple sets of teaching data into the learning model, each set of teaching data of the multiple sets of teaching data including sample behavior data for identifying the historical records of behaviors performed by a sample user corresponding to the group and sample sleep state data for identifying the sleep state of a sample user corresponding to the group."

[0454] The computer program involved in the third mode is capable of "enabling the at least one processor to perform the following functions in the above-mentioned first mode: connecting to the inference model generated by inputting multiple sets of teaching data into the learning model via a communication line, each set of teaching data of the multiple sets of teaching data including sample behavior data for identifying the historical records of behaviors performed by a sample user corresponding to the group and sample sleep state data for identifying the sleep state of a sample user corresponding to the group."

[0455] The computer program involved in the fourth mode is capable of "enabling the at least one processor to perform the following functions in the above-mentioned first mode: obtaining multiple sets of teaching data, generating the inference model by inputting the multiple sets of teaching data into a learning model and making it learn, each set of teaching data of the multiple sets of teaching data including sample behavior data for identifying the historical records of behaviors performed by a sample user corresponding to the group and sample sleep state data for identifying the sleep state of a sample user corresponding to the group."

[0456] The computer program involved in the fifth mode is capable of "causing the at least one processor to perform the following functions in the above-mentioned fourth mode: the plurality of sets of teaching data obtained include first plurality of sets of teaching data, each set of teaching data included in the first plurality of sets of teaching data includes at least sample property data for identifying the sleep property of a sample user corresponding to the group and sample rhythm data for identifying the sleep rhythm of a sample user corresponding to the group as the sample sleep state data, performing preprocessing on each set of teaching data included in the first plurality of sets of teaching data obtained, and inputting the preprocessed first plurality of sets of teaching data into the learning model to generate the inference model , outputting object sleep type data for identifying the sleep type of the object user from the inference model as the object sleep state data, in the preprocessing, based on the sample property data for identifying the sleep property of a sample user corresponding to the group and the sample rhythm data for identifying the sleep rhythm of a sample user corresponding to the group, generating sample sleep type data for identifying the sleep type of a sample user corresponding to the group, each group of teaching data of the first plurality of groups of teaching data after the preprocessing includes the sample behavior data of a sample user corresponding to the group and the sample sleep type data of a sample user corresponding to the group.

[0457] The computer program involved in the sixth mode is capable of "enabling the at least one processor to perform the following functions in any of the above-mentioned second to fourth modes: the multiple groups of teaching data include second multiple groups of teaching data, and each group of teaching data included in the second multiple groups of teaching data includes at least sample sleep type data for identifying the sleep type of a sample user corresponding to the group, which is determined based on sample property data for identifying the sleep property of a sample user corresponding to the group and sample rhythm data for identifying the sleep rhythm of a sample user corresponding to the group, as the sample sleep state data, and object sleep type data for identifying the sleep type of the object user is output from the inference model as the object sleep state data."

[0458] Regarding the computer program involved in the 7th method, in the above-mentioned 5th method or the above-mentioned 6th method, it can be "the sample property data for identifying the sleep property of a sample user corresponding to the group represents a score calculated based on the sample user's answer to the Athens Insomnia Scale, the Pittsburgh Sleep Questionnaire, the Three-Dimensional Sleep Scale or the Insomnia Severity Questionnaire, the sample rhythm data for identifying the sleep rhythm of a sample user corresponding to the group represents the absolute value of the difference between the central time of the sample user's usual sleep time period and the central time period of the sleep time period on weekends, that is, the social jet lag, and the sleep type to which the sample user corresponding to the group belongs is one of a plurality of categories divided according to at least one threshold value set in association with the score on one of the horizontal axis and the vertical axis and at least one threshold value set in association with the social jet lag on the other of the horizontal axis and the vertical axis."

[0459] The computer program involved in the 8th mode is capable of "enabling the at least one processor to perform the following functions in any of the above-mentioned 2nd mode to the above-mentioned 4th mode: the multiple groups of teaching data include a third plurality of groups of teaching data, each group of teaching data included in the third plurality of groups of teaching data includes at least sample property data for identifying the sleep property of a sample user corresponding to the group and sample rhythm data for identifying the sleep rhythm of a sample user corresponding to the group as the sample sleep state data, and outputs object property data for identifying the sleep property of the target user and object rhythm data for identifying the sleep rhythm of the target user from the inference model as the target sleep state data."

[0460] The computer program involved in the 9th mode is capable of "enabling the at least one processor to perform the following functions in any of the above-mentioned methods 2 to 4: the multiple groups of teaching data include fourth multiple groups of teaching data, each group of teaching data included in the fourth multiple groups of teaching data includes at least sample property data for identifying the sleep properties of a sample user corresponding to the group as the sample sleep state data, and object property data for identifying the sleep properties of the object user is output from the inference model as the object sleep state data."

[0461] The computer program involved in the tenth mode is capable of "enabling the at least one processor to perform the following functions in any of the above-mentioned second to fourth modes: the multiple groups of teaching data include fifth multiple groups of teaching data, each group of teaching data included in the fifth multiple groups of teaching data includes at least sample rhythm data for identifying the sleep rhythm of a sample user corresponding to the group as the sample sleep state data, and object rhythm data for identifying the sleep rhythm of the object user is output from the inference model as the object sleep state data."

[0462] The computer program involved in the 11th method is capable of "enabling the at least one processor to perform the following functions in any one of the above-mentioned methods 2 to 4: the multiple groups of teaching data include sixth multiple groups of teaching data, each group of teaching data included in the sixth multiple groups of teaching data includes at least sample time data for identifying the sleep time of a sample user corresponding to the group as the sample sleep state data, and object time data for identifying the sleep time of the object user is output from the inference model as the object sleep state data."

[0463] The computer program involved in the twelfth embodiment is capable of "causing the at least one processor to perform the following function: determining at least one target product and / or at least one target service corresponding to the target sleep state data from a plurality of products and / or a plurality of services." in any one of the first to eleventh embodiments.

[0464] The computer program involved in the 13th method is capable of "enabling the at least one processor to perform the following function: by inputting the object sleep state data into another inference model generated by performing supervised learning, outputting recommendation data for identifying the at least one object product and / or the at least one object service from the other inference model."

[0465] The computer program involved in the 14th method is capable of "enabling the at least one processor to perform the following function: by inputting the object sleep state data into a retrieval table obtained by establishing associations between the object sleep state data and the multiple products and / or the multiple services, obtaining recommendation data for identifying the at least one object product and / or the at least one object service from the retrieval table."

[0466] The computer program according to the fifteenth aspect may be "the at least one processor includes a central processing unit (CPU), a microprocessor and / or a graphics processing unit (GPU)" in any one of the first to fourteenth aspects.

[0467] The information processing device involved in the 16th mode can be "configured as: comprising at least one processor, the at least one processor acquiring object behavior data for identifying the historical records of behaviors performed by the object user, and outputting object sleep state data for identifying the sleep state of the object user from the inference model by inputting the object behavior data into an inference model generated by performing supervised learning, the object sleep state data including at least one of object property data for identifying the sleep property of the object user, object rhythm data for identifying the sleep rhythm of the object user, object time data for identifying the sleep time of the object user, and object type data for identifying the sleep type to which the object user belongs."

[0468] The information processing device according to the seventeenth aspect can be "a terminal device or a server device" in the sixteenth aspect.

[0469] Regarding the information processing device involved in the 18th embodiment, in the above-mentioned 17th embodiment, it can be "the at least one processor includes a central processing unit (CPU), a microprocessor and / or a graphics processing unit (GPU)".

[0470] The method involved in the 19th embodiment is capable of "being a method executed by at least one processor that executes commands that can be read by a computer, comprising the following stages: the at least one processor obtains object behavior data for identifying the historical record of behaviors performed by the object user by executing the commands; and by inputting the object behavior data into an inference model generated by performing supervised learning, outputting object sleep state data for identifying the sleep state of the object user from the inference model, the object sleep state data including at least one of object property data for identifying the sleep property of the object user, object rhythm data for identifying the sleep rhythm of the object user, object time data for identifying the sleep time of the object user, and object type data for identifying the sleep type to which the object user belongs."

[0471] Regarding the method involved in the 20th embodiment, in the above-mentioned 19th embodiment, it can be "the at least one processor includes a central processing unit (CPU), a microprocessor and / or a graphics processing unit (GPU)."

[0472] The method involved in the 21st embodiment is capable of "being a method executed by at least one processor that executes commands that can be read by a computer, comprising the following stages: an acquisition stage, in which the at least one processor acquires multiple sets of teaching data by executing the commands, each set of teaching data of the multiple sets of teaching data includes sample behavior data for identifying the historical records of behaviors performed by a sample user corresponding to the group and sample sleep state data for identifying the sleep state of a sample user corresponding to the group; and a generation stage, in which an inference model is generated by inputting the multiple sets of teaching data into a learning model and making it learn, the inference model being configured to receive object behavior data for identifying the historical records of behaviors performed by an object user as input, and output object sleep state data for identifying the sleep state of the object user, the object sleep state data including at least one of object property data for identifying the sleep property of the object user, object rhythm data for identifying the sleep rhythm of the object user, object time data for identifying the sleep time of the object user, and object type data for identifying the sleep type to which the object user belongs."

[0473] Regarding the method involved in the 22nd embodiment, in the above-mentioned 21st embodiment, it can be "the acquisition stage includes a stage of acquiring the first plurality of groups of teaching data included in the plurality of groups of teaching data, each group of teaching data included in the first plurality of groups of teaching data at least includes sample property data for identifying the sleep properties of a sample user corresponding to the group and sample rhythm data for identifying the sleep rhythm of a sample user corresponding to the group as the sample sleep state data, and the generation stage includes the following stages: for each group of teaching data included in the first plurality of groups of teaching data acquired, performing preprocessing such as generating sample sleep type data for identifying the sleep type of a sample user corresponding to the group based on the sample property data for identifying the sleep properties of a sample user corresponding to the group and the sample rhythm data for identifying the sleep rhythm of a sample user corresponding to the group; and inputting the preprocessed first plurality of groups of teaching data into the learning model to generate the inference model, each group of teaching data of the preprocessed first plurality of groups of teaching data includes the sample behavior data of a sample user corresponding to the group and the sample sleep type data of a sample user corresponding to the group".

[0474] Regarding the method involved in the 23rd mode, in the above-mentioned 21st mode, it can be "the acquisition stage includes a stage of acquiring second multiple groups of teaching data included in the multiple groups of teaching data, each group of teaching data included in the second multiple groups of teaching data includes at least sample sleep type data for identifying the sleep type of a sample user corresponding to the group determined based on sample property data for identifying the sleep property of a sample user corresponding to the group and sample rhythm data for identifying the sleep rhythm of a sample user corresponding to the group, as the sample sleep state data, and the method includes a stage of inputting the second multiple groups of teaching data into the learning model to generate the inference model."

[0475] Regarding the method involved in the 24th embodiment, in the above-mentioned 23rd embodiment, "the sample property data for identifying the sleep property of a sample user corresponding to the group represents a score calculated by the sample user's answer to the Athens Insomnia Scale, the Pittsburgh Sleep Questionnaire, the Three-Dimensional Sleep Scale or the Insomnia Severity Questionnaire, the sample rhythm data for identifying the sleep rhythm of a sample user corresponding to the group represents the absolute value of the difference between the central time of the sample user's usual sleep time period and the central time period of the sleep time period on weekends, namely, the social jet lag, and the sleep type to which the sample user corresponding to the group belongs is one of a plurality of categories divided according to at least one threshold value set in association with the score on one of the horizontal and vertical axes and at least one threshold value set in association with the social jet lag on the other of the horizontal and vertical axes."

[0476] Regarding the method involved in Mode 25, any one of Modes 21 to 24 may "also include an inference stage, in which object sleep state data for identifying the sleep state of the object user is output from the inference model by inputting object behavior data of a historical record for identifying behaviors performed by the object user into the inference model."

[0477] Regarding the method involved in the 26th mode, in any of the above-mentioned 21st mode to the above-mentioned 25th mode, it can be "the at least one processor includes a central processing unit (CPU), a microprocessor and / or a graphics processing unit (GPU)".

[0478] The computer program involved in the 27th method is capable of "being executed by at least one processor so that the at least one processor performs the following functions: obtaining object behavior data for identifying the historical records of behaviors performed by the object user, and outputting object menopausal state data for identifying the state of menopausal symptoms of the object user from the inference model by inputting the object behavior data into an inference model generated by performing supervised learning, the object menopausal state data including at least one of object comprehensive state data for identifying the comprehensive state of the menopausal symptoms of the object user, object first state data for identifying the state of a first symptom of the object user, and object second state data for identifying the state of a second symptom of the object user."

[0479] The computer program involved in the 28th method is capable of "enabling the at least one processor to perform the following function in the above-mentioned 27th method: obtaining the inference model generated by inputting multiple groups of teaching data into the learning model, each group of teaching data of the multiple groups of teaching data includes sample behavior data for identifying the historical record of behavior performed by a sample user corresponding to the group and sample menopausal status data for identifying the status of menopausal symptoms of a sample user corresponding to the group."

[0480] The computer program involved in the 29th method is capable of "enabling the at least one processor to perform the following functions in the above-mentioned 27th method: connecting to the inference model generated by inputting multiple sets of teaching data into the learning model via a communication line, each set of teaching data of the multiple sets of teaching data includes sample behavior data for identifying the historical records of behaviors performed by a sample user corresponding to the group and sample menopausal status data for identifying the status of menopausal symptoms of a sample user corresponding to the group."

[0481] The computer program involved in the 30th method is capable of "enabling the at least one processor to perform the following functions in the above-mentioned 27th method: obtaining multiple sets of teaching data, generating the inference model by inputting the multiple sets of teaching data into the learning model and making it learn, each set of teaching data of the multiple sets of teaching data includes sample behavior data for identifying the historical records of behaviors performed by a sample user corresponding to the group and sample menopausal status data for identifying the status of menopausal symptoms of a sample user corresponding to the group."

[0482] The computer program involved in the 31st method is capable of "enabling the at least one processor to perform the following functions in the above-mentioned 30th method: the multiple groups of teaching data obtained include first multiple groups of teaching data, each group of teaching data included in the first multiple groups of teaching data includes at least sample comprehensive state data for identifying the comprehensive state of menopausal symptoms of a sample user corresponding to the group as the sample menopausal state data, and object comprehensive state data for identifying the comprehensive state of menopausal symptoms of the object user is output from the inference model as the object menopausal state data."

[0483] The computer program involved in the 32nd method is capable of "enabling the at least one processor to perform the following functions in the above-mentioned 31st method: each group of teaching data included in the first plurality of groups of teaching data at least includes sample comprehensive status data of a score corresponding to the comprehensive status of menopausal symptoms of a sample user corresponding to the group as the sample menopausal status data, preprocessing is performed on each group of teaching data included in the first plurality of groups of teaching data, the first plurality of groups of teaching data after the preprocessing are input into the learning model to generate the inference model, and object comprehensive status data for identifying the category corresponding to the comprehensive status of menopausal symptoms of the object user is output from the inference model as the object menopausal status data, and in the preprocessing, the sample comprehensive status data of the score corresponding to the comprehensive status of menopausal symptoms of a sample user corresponding to the group is transformed into sample comprehensive status data of a category corresponding to the comprehensive status of menopausal symptoms of a sample user corresponding to the group."

[0484] Regarding the computer program involved in the 33rd method, in the above-mentioned 32nd method, it can be "a score corresponding to the comprehensive state of menopausal symptoms of a sample user corresponding to the group is calculated based on the answers of a sample user corresponding to the group to the Simplified Menopause Index (SMI), Menopause Rating Scale, Kupperman Index, Green Climacteric Scale, PSST (Premenstrual Symptoms Screening Tool), WHQ (Women's Health Questionnaire), VAS (Visual Analogue Scale), HFRDI (Hot Flash Related Daily Interference Scale), HFCS (Hot Flash Composite Score), MENQOL (Menopause-Specific Quality Of Life) or Japanese Women's Menopausal Symptom Evaluation Form."

[0485] The computer program involved in the 34th method is capable of "enabling the at least one processor to perform the following functions in the above-mentioned 30th method: the multiple groups of teaching data obtained include a second plurality of groups of teaching data, each group of teaching data included in the second plurality of groups of teaching data includes at least sample first state data for identifying the state of a first symptom of a sample user corresponding to the group as the sample menopausal state data, and object first state data for identifying the state of the first symptom of the object user is output from the inference model as the object menopausal state data."

[0486] The computer program involved in the 35th method is capable of "enabling the at least one processor to perform the following functions in the above-mentioned 34th method: each group of teaching data included in the second plurality of groups of teaching data at least includes sample first state data of a score corresponding to the state of a first symptom of a sample user corresponding to the group as the sample menopausal state data, preprocessing is performed on each group of teaching data included in the second plurality of groups of teaching data, the second plurality of groups of teaching data after the preprocessing are input into the learning model to generate the inference model, and object first state data for identifying the category corresponding to the state of the first symptom of the object user is output from the inference model as the object menopausal state data, and in the preprocessing, the sample first state data of the score corresponding to the state of the first symptom of a sample user corresponding to the group is transformed into sample first state data of a category corresponding to the state of the first symptom of a sample user corresponding to the group."

[0487] The computer program involved in the 36th method is capable of "enabling the at least one processor to perform the following functions in the above-mentioned 30th method: the multiple groups of teaching data obtained include a third plurality of groups of teaching data, each group of teaching data included in the third plurality of groups of teaching data includes at least sample second state data for identifying the state of a second symptom of a sample user corresponding to the group as the sample menopausal state data, and object first state data for identifying the state of the second symptom of the object user is output from the inference model as the object menopausal state data."

[0488] The computer program involved in the 37th method is capable of "enabling the at least one processor to perform the following functions in the above-mentioned 36th method: each group of teaching data included in the third plurality of groups of teaching data at least includes sample second state data of a score corresponding to the state of the second symptom of a sample user corresponding to the group as the sample menopausal state data, performs preprocessing on each group of teaching data included in the third plurality of groups of teaching data, inputs the third plurality of groups of teaching data after the preprocessing into the learning model to generate the inference model, and outputs object second state data of a category corresponding to the state of the second symptom of the object user from the inference model as the object menopausal state data, and in the preprocessing, transforms the sample second state data of a score corresponding to the state of the second symptom of a sample user corresponding to the group into sample second state data of a category corresponding to the state of the second symptom of a sample user corresponding to the group."

[0489] The computer program involved in the 38th mode is capable of "causing the at least one processor to perform the following function: determining at least one target product and / or at least one target service corresponding to the target menopausal status data from a plurality of products and / or a plurality of services" in any one of the above-mentioned 27th mode to the above-mentioned 37th mode.

[0490] The computer program involved in the 39th embodiment is capable of "enabling the at least one processor to perform the following function: by inputting the target menopausal status data into another inference model generated by performing supervised learning, outputting recommendation data for identifying the at least one target product and / or the at least one target service from the other inference model."

[0491] The computer program involved in the 40th method is capable of "enabling the at least one processor to perform the following function in the above-mentioned 38th method: by inputting the target menopausal status data into a retrieval table obtained by establishing an association between the target menopausal status data and the multiple products and / or the multiple services, obtaining recommendation data for identifying the at least one target product and / or the at least one target service from the retrieval table."

[0492] Regarding the computer program involved in the 41st mode, in any of the above-mentioned 27th mode to the above-mentioned 40th mode, it can be "the at least one processor includes a central processing unit (CPU), a microprocessor and / or a graphics processing unit (GPU)".

[0493] The information processing device involved in the 42nd method is capable of "being configured to have at least one processor, which obtains object behavior data for identifying the historical records of behaviors performed by the object user, and outputs object menopausal state data for identifying the state of menopausal symptoms of the object user from the inference model by inputting the object behavior data into an inference model generated by performing supervised learning, wherein the object menopausal state data includes at least one of object comprehensive state data for identifying the comprehensive state of the menopausal symptoms of the object user, object first state data for identifying the state of a first symptom of the object user, and object second state data for identifying the state of a second symptom of the object user."

[0494] The information processing device according to the 43rd aspect can be "a terminal device or a server device" in the above-mentioned 42nd aspect.

[0495] Regarding the information processing device involved in the 44th embodiment, in the above-mentioned 42nd embodiment or the above-mentioned 43rd embodiment, it can be "the at least one processor includes a central processing unit (CPU), a microprocessor and / or a graphics processing unit (GPU)".

[0496] The method involved in the 45th embodiment is capable of "being a method executed by at least one processor that executes commands that can be read by a computer, comprising the following stages: the at least one processor obtains object behavior data for identifying the historical record of behaviors performed by the object user by executing the commands; and by inputting the object behavior data into an inference model generated by performing supervised learning, outputting object menopausal state data for identifying the state of menopausal symptoms of the object user from the inference model, the object menopausal state data including at least one of object comprehensive state data for identifying the comprehensive state of the menopausal symptoms of the object user, object first state data for identifying the state of a first symptom of the object user, and object second state data for identifying the state of a second symptom of the object user."

[0497] Regarding the method involved in the 46th embodiment, in the above-mentioned 45th embodiment, it can be "the at least one processor includes a central processing unit (CPU), a microprocessor and / or a graphics processing unit (GPU)."

[0498] The method involved in the 47th mode is capable of "being a method executed by at least one processor that executes commands that can be read by a computer, comprising the following stages: an acquisition stage, in which the at least one processor acquires multiple sets of teaching data by executing the commands, each set of teaching data of the multiple sets of teaching data includes sample behavior data for identifying the historical records of behaviors performed by a sample user corresponding to the group and sample menopausal status data for identifying the state of menopausal symptoms of a sample user corresponding to the group; and a generation stage, in which an inference model is generated by inputting the multiple sets of teaching data into a learning model and making it learn, the inference model being configured to be input with object behavior data for identifying the historical records of behaviors performed by an object user, and outputting object menopausal status data for identifying the state of menopausal symptoms of the object user, the object menopausal status data including at least one of object comprehensive status data for identifying the comprehensive state of menopausal symptoms of the object user, object first status data for identifying the state of a first symptom of the object user, and object second status data for identifying the state of a second symptom of the object user".

[0499] Regarding the method involved in the 48th embodiment, in the above-mentioned 47th embodiment, it can be "the acquisition stage includes a stage of acquiring the first plurality of groups of teaching data included in the plurality of groups of teaching data, each group of teaching data included in the first plurality of groups of teaching data at least includes sample comprehensive status data of a score corresponding to the comprehensive status of menopausal symptoms of a sample user corresponding to the group as the sample menopausal status data, and the generation stage includes the following stages: performing the following preprocessing on each group of teaching data included in the first plurality of groups of teaching data obtained, in which the sample comprehensive status data of the score corresponding to the comprehensive status of menopausal symptoms of a sample user corresponding to the group is transformed into sample comprehensive status data of a category corresponding to the comprehensive status of menopausal symptoms of a sample user corresponding to the group; and inputting the preprocessed first plurality of groups of teaching data into the learning model to generate the inference model, each group of teaching data of the preprocessed first plurality of groups of teaching data includes the sample behavior data of a sample user corresponding to the group and the sample comprehensive status data of a sample user corresponding to the group".

[0500] Regarding the method involved in the 49th method, in the above-mentioned 47th method, it can be "the acquisition step includes a stage of acquiring a second plurality of groups of teaching data included in the plurality of groups of teaching data, each group of teaching data included in the second plurality of groups of teaching data at least includes sample first state data of a score corresponding to the state of a first symptom of a sample user corresponding to the group as the sample menopausal state data, and the generation stage includes the following stages: performing the following preprocessing on each group of teaching data included in the acquired second plurality of groups of teaching data, in which the sample first state data of the score corresponding to the state of the first symptom of a sample user corresponding to the group is transformed into sample first state data of a category corresponding to the state of the first symptom of a sample user corresponding to the group; and inputting the preprocessed second plurality of groups of teaching data into the learning model to generate the inference model, each group of teaching data of the preprocessed second plurality of groups of teaching data including the sample behavior data of a sample user corresponding to the group and the sample first state data of a sample user corresponding to the group".

[0501] Regarding the method involved in the 50th embodiment, in the above-mentioned 47th embodiment, it can be "the acquisition stage includes a stage of acquiring a third plurality of groups of teaching data included in the plurality of groups of teaching data, each group of teaching data included in the third plurality of groups of teaching data at least includes sample second state data of a score corresponding to the state of a second symptom of a sample user corresponding to the group as the sample menopausal state data, and the generation stage includes the following stages: performing the following preprocessing on each group of teaching data included in the acquired third plurality of groups of teaching data, in which the sample second state data of the score corresponding to the state of a second symptom of a sample user corresponding to the group is transformed into sample second state data of a category corresponding to the state of the second symptom of a sample user corresponding to the group; and inputting the preprocessed third plurality of groups of teaching data into the learning model to generate the inference model, each group of teaching data of the preprocessed third plurality of groups of teaching data including the sample behavior data of a sample user corresponding to the group and the sample second state data of a sample user corresponding to the group".

[0502] Regarding the method involved in the 51st mode, in any of the above-mentioned 47th mode to the above-mentioned 50th mode, it can be "the at least one processor includes a central processing unit (CPU), a microprocessor and / or a graphics processing unit (GPU)".

[0503] The computer program involved in the 52nd method is capable of "being executed by at least one processor so as to cause the at least one processor to perform the following functions: obtaining object behavior data for identifying the historical records of behaviors performed by the object user, and outputting state data for identifying the exercise state, the state associated with lifestyle diseases, or the state of insufficient physical activity of the object user from the inference model by inputting the object behavior data into an inference model generated by performing supervised learning, wherein the state data includes at least one of data for identifying the weight of the object user, data for identifying the BMI (Body Mass Index) of the object user, data for identifying the number of steps of the object user, data for identifying whether the object user smokes, and data for identifying a metabolic syndrome diagnostic benchmark value of the object user."

[0504] The computer program involved in the fifty-third embodiment is capable of "being executed by at least one processor to cause the at least one processor to perform the following functions: obtaining object behavior data for identifying a history of behaviors performed by a target user, and outputting status data for identifying a hydration status of the target user from the inference model by inputting the object behavior data into an inference model generated by performing supervised learning, wherein the status data includes at least one of data for identifying a serum sodium value of the target user, data for identifying a urine osmotic pressure of the target user, data for identifying a urine specific gravity of the target user, data for identifying a urine color of the target user, data for identifying a BUN (urea nitrogen) / creatinine ratio of the target user, data for identifying a dehydration evaluation scale of the target user, and data for identifying subjective symptoms associated with insufficient water of the target user."

[0505] The computer program involved in the 54th method is capable of "being executed by at least one processor so as to cause the at least one processor to perform the following functions: obtaining object behavior data for identifying the historical records of behaviors performed by the object user, and outputting status data for identifying the immune status of the object user from the inference model by inputting the object behavior data into an inference model generated by performing supervised learning, wherein the status data includes at least one of data for identifying the susceptibility of the object user to colds, data for identifying the frequency of colds of the object user, data for identifying the SIgA concentration of the object user, data for identifying the allergic symptoms of the object user, data for identifying the status of the oral environment and / or immunity of the object user, data for identifying the sleep status of the object user, data for identifying the exercise status of the object user, and data for identifying the stress status of the object user."

[0506] The computer program involved in the 55th method is capable of "being executed by at least one processor so as to cause the at least one processor to perform the following functions: obtaining object behavior data for identifying the historical records of behaviors performed by the object user, and outputting status data for identifying the nutritional status of the object user from the inference model by inputting the object behavior data into an inference model generated by performing supervised learning, the status data including at least one of data for identifying the presence or absence of subjective symptoms caused by nutritional decline of the object user, data for identifying the diversity score (DVS) of the food intake of the object user, data for identifying the shortened version of the appetite scale (Simplified Nutritional Appetite Questionnaire) of the object user, and data for identifying the BMI (Body Mass Index) of the object user."

[0507] II. Chapter 2

[0508] The techniques described in Chapter 1 and the techniques described in Chapter 2 as follows can be used in combination.

[0509] For example, in the technology to be described below, the inference model for inferring the sleep state described in the first chapter can be used as the sleep state prediction model 110b. Similarly, in the technology to be described below, the inference model for inferring the state of menopausal symptoms described in the first chapter can be used as the female health prediction model 100a. And, similarly, in the technology to be described below, the inference model for predicting the exercise state, the inference model for predicting the immune state, the inference model for predicting the water replenishment state, and the inference model for predicting the nutritional state described in the first chapter can be used as the lifestyle disease prediction model 110c, the immune state prediction model 110e, the water state prediction model 110d, and the nutritional state prediction model 110f, respectively.

[0510] Without limitation to this, at least one feature included in the technology described in Chapter 1 (or at least one feature included in the technology to be described in Chapter 2) can also be utilized in the technology to be described in Chapter 2 (or the technology described in Chapter 1).

[0511] In recent years, the living environment has been changing significantly around the world due to population decline, economic downturn, natural disasters, and the spread of infectious diseases. On the other hand, opportunities for effective use of artificial intelligence (AI) are increasing, and the development and popularization of the fifth-generation mobile communication system are advancing. As the living environment is changing, the public infrastructure that supports society has begun to shift to an industrial framework that is related to data based on the effective use of technology, and social activities and behavior information have begun to be visualized.

[0512] In the process of increasing attention to health, it is required that the response to health management also adapts to social trends. So far, it has mainly focused on "passive medical care" in which patients are treated after symptoms begin to worsen. However, it is believed that "advance medical care" and "preventive medical care" will be carried out in the future. Among them, "advance medical care" refers to healthy people obtaining their own data at home, collecting the data and digitizing each person's health status to foresee future health management, and "preventive medical care" refers to disease prevention. Moreover, it is also possible to imagine "predictive medical care" that predicts future diseases and calls attention to health.

[0513] Patent Document 1 discloses a method of constructing an inference model for inferring the health status of a target user based on data describing the purchase history of the target user. This patent document is hereby incorporated by reference in its entirety into this specification.

[0514] There are various risks in health status, but the method described in Patent Document 1 does not take into account the fact that customers understand various health risks.

[0515] Several aspects of the present invention have been made in view of any of the above-mentioned problems, and provide a program, an information processing device, and an information processing method that can appropriately provide health-related information to customers.

[0516] The program involved in an exemplary embodiment of the present invention is used to enable an information processing device to perform the following processing: obtaining first purchase data representing the purchase history of a target customer; prediction processing, using a health prediction model related to multiple health symptoms from the first to the Nth (N is an integer greater than 2) to predict the degree of the multiple health symptoms based on the first purchase data; and providing health information representing the predicted degrees of the multiple health symptoms to the target customer, wherein the health prediction model related to the kth (1≤k≤N) health symptom calculates a numerical value representing the degree of the kth health symptom based on the first purchase data related to the product category.

[0517] An information processing device involved in an exemplary embodiment of the present invention comprises: an input unit, which acquires first purchase data representing a purchase history of a target customer; a prediction unit, which predicts the extent of the multiple health symptoms using a health prediction model related to multiple health symptoms from the first to the Nth (N is an integer greater than 2) types based on the first purchase data; and an output unit, which provides the target customer with health information indicating the predicted extent of the multiple health symptoms, wherein the health prediction model related to the kth (1≤k≤N) health symptom calculates a numerical value indicating the extent of the kth health symptom based on the first purchase data related to the product category.

[0518] The information processing method involved in an exemplary embodiment of the present invention is an information processing method in which an information processing device performs the following processing: obtaining first purchase data representing the purchase history of a target customer; predicting the extent of the multiple health symptoms using a health prediction model related to multiple health symptoms from the 1st to the Nth (N is an integer greater than 2) types based on the first purchase data; and providing health information indicating the predicted extent of the multiple health symptoms to the target customer, wherein the health prediction model related to the kth (1≤k≤N) health symptom of the information processing device calculates a numerical value indicating the extent of the kth health symptom based on the first purchase data related to the product category.

[0519] Furthermore, in the present invention, "part", "component", "device", and "system" refer not only to physical parts, but also include cases where the functions of the "part", "component", "device", and "system" are realized by software. In addition, the functions of one "part", "component", "device", and "system" may be realized by two or more physical parts or devices, and the functions of two or more "parts", "components", "devices", and "systems" may be realized by one physical part or device.

[0520] 1 Summary

[0521] Figures 15 to 28 This is a diagram for explaining the implementation method. Below, it is roughly divided into a health prediction system for predicting a customer's health status (also called health symptoms) and a learning system for generating a prediction model for predicting a customer's health status used in the health prediction system, and the overview, functional structure, processing flow, etc. are explained in turn.

[0522] 2. Health prediction system

[0523] 2.1 Overview of the Processing of the Health Prediction System

[0524] First, refer to Fig.15 An outline of a rough process of the health prediction system 1Z according to the embodiment will be described. Fig.15 This is a diagram schematically showing the flow of data for performing health prediction of a customer using the health prediction system 1Z according to the embodiment.

[0525] The health prediction system 1Z is a system for inferring the health status of an individual based on the individual's purchasing behavior. If the health prediction system 1Z is used, the health status of the individual customer can be inferred based on the customer's purchasing behavior, i.e., the purchase data, in the absence of data related to the individual's health, such as medical examination data. The inferred health status can, for example, help improve the health of the customer.

[0526] Here, the comprehensive health prediction program 100Z for predicting the health status of an individual based on the individual's purchasing behavior includes a plurality of learned prediction models for predicting the health status based on the individual's purchasing behavior. Fig.15 In the example, the comprehensive health prediction program 100Z is constructed by combining six prediction models, but is not limited to six, and may be less than five or more than seven.

[0527] The comprehensive health prediction program 100Z involved in this embodiment includes six prediction models (hereinafter collectively referred to as "prediction models 110Z"), including a female health prediction model 110a for predicting female health, a sleep state prediction model 110b for predicting sleep state, a lifestyle disease prediction model 110c for predicting lifestyle diseases, a water state prediction model 110d for predicting water state, an immune state prediction model 110e for predicting immune state, and a nutritional state prediction model 110f for predicting nutritional state. The prediction models included in the comprehensive health prediction program 100Z calculate values ​​(values ​​indicating the possibility and / or degree of suffering from a specific disease) representing female health states (SMI related to menopausal symptoms, etc.) / sleep states / lifestyle diseases / water states (hydration replenishment status) / immune states / nutritional states based on purchase data 150Z representing the customer's purchase history as the customer's health prediction value 160Z. Fig.15 In the example, 6 values ​​representing health status are shown in %. Depending on the type of disease, only males can be set as objects, only females can be set as objects, or both males and females can be set as objects. These prediction models 110Z can be described as functions. These prediction models 110Z are constructed by machine learning using teaching data including purchase history records of products and various health-related statuses. The method of constructing the prediction model is described later.

[0528] More specifically, the comprehensive health prediction program 100Z receives the customer's purchase data 150Z. The purchase data 150Z is data related to the purchase history of the customer whose health is to be predicted, and is data corresponding to the explanatory variables of the teaching data when constructing each prediction model 110Z included in the comprehensive health prediction program 100Z. The comprehensive health prediction program 100Z inputs the purchase data 150Z to each prediction model 110Z in parallel, and calculates the customer's health prediction value 160Z. In addition, in the present embodiment, only the purchase data 150Z is used as an input to the prediction model 110Z, but for example, the customer's attribute information such as the customer's gender and age may also be input.

[0529] In addition, the comprehensive health prediction program 100Z can also provide various suggestions to customers regarding health improvement based on the calculated health prediction value 160Z. More specifically, for example, consider indicating to customers categories with high health prediction values ​​and suggesting measures that can be taken when the customer has not noticed (potential) symptoms or when the symptoms that are currently appearing are mild. For more information about this suggestion method, see Fig.18 The following will be described later.

[0530] 2.2 Structure of the comprehensive health prediction program

[0531] Below, refer to Fig.16 To illustrate the functional structure of the comprehensive health prediction program 100Z. Fig.16 1 is a diagram showing the functional structure of the comprehensive health prediction program 100Z. The comprehensive health prediction program 100Z includes an input unit 120Z, a product category determination unit 125Z, a prediction model 110Z, an output unit 130Z, and a suggestion unit 135Z. In addition, the comprehensive health prediction program 100Z can be installed as a single program or as multiple programs that can operate collaboratively through data input and output.

[0532] The input unit 120Z receives (acquires) input of the purchase data 150Z. The input unit 120Z may read the purchase data 150Z from a storage medium such as a hard disk drive (HDD) or a solid state drive (SSD) built into an information processing device (computer) executing the comprehensive health prediction program 100Z, or may receive input of the purchase data 150Z from an external information processing device connected via a LAN (Local Area Network), the Internet, or the like.

[0533] Here, refer to Fig.17 A specific example of purchase data 150Z inputted by input unit 120Z will be described. Fig.17 1 is a diagram showing a specific example of purchase data 150Z. Here, purchase data 150Z can be panel data storing the purchase history of customers provided by a survey company. The panel data is data obtained by scanning a product when the customer, as the purchaser, purchases the product. When a customer uses his or her own smartphone or a loaned barcode reader to scan a product, the product is stored in association with the pre-set customer attributes (company employee / student / housewife, etc.), purchase target information, etc. This makes it possible to understand what kind of person purchased what kind of thing in what kind of store.

[0534] Fig.17The example of purchase data 150Z shows the purchase history of food and / or daily necessities purchased by a customer within a certain period (for example, one year). Each commodity purchased by the customer is classified by commodity category in the JICFS (JAN Item Code File Service) classification. In other words, at least one commodity belongs to each commodity category. In the purchase data 150Z, each item 150a of about 500 varieties (for example: soft drinks, mouthwash, etc.) of the fine categories included in the major categories "food" and "daily necessities" in the JICFS classification related to the commodity category to which the commodity purchased by the customer belongs is input in association with the numerical data such as the purchase amount, purchase quantity, and number of purchases during the period, i.e., the performance data 150b, according to each customer ID.

[0535] In addition, in the present embodiment, the purchase data 150Z includes the purchase history records of food and / or daily necessities, but it is not limited to these purchase history records as long as the association with the health status of the customer can be inferred. For example, it is also possible to include the purchase history records of other commodities such as medicines, massage, acupuncture treatment and other services in the purchase data 150Z. In addition, it is also possible to determine the purchase data 150Z based on the settlement information managed by the retail store, and it is also possible to determine the purchase data 150Z using, for example, the settlement information of the credit card, the payment information to the financial institution, etc.

[0536] When receiving the purchase data 150Z from the input unit 120Z, the commodity category determination unit 125Z determines the classification code (JICFS classification) and commodity category input to each prediction model 110Z based on the purchase data 150Z input to each prediction model 110Z, and inputs numerical data such as the purchase amount, purchase quantity, and purchase frequency related to the obtained classification code and commodity category, namely, performance data 150b (only part of them may be used) into each prediction model 110Z in parallel.

[0537] The classification code (product category) of the purchase data 150Z input to the prediction model 110Z can be determined for each prediction model 110Z. The method of determining the product category as input to the prediction model 110Z will be described later in 3.4 below regarding the creation of the prediction model 110Z.

[0538] Furthermore, it is also possible to input the purchase data 150Z input from the input unit 120Z directly into the prediction model 110Z without specifically limiting the purchase data 150Z by the commodity category. In this case, the commodity category identification unit 125Z is unnecessary.

[0539] The prediction model 110Z receives the input of the numerical data 150b, i.e., the purchase amount, purchase quantity, and purchase frequency, related to the classification code and the product category selected by the product category determination unit 125Z, to calculate the health prediction values ​​of women's health status (menopausal symptoms), sleep status (sleep disorders), lifestyle diseases (lack of physical activity), water status (degree of water deficiency / water replenishment status), immune status (degree of immune decline), and nutritional status (degree of nutritional deficiency). Fig.15 In the example, for the customer with user ID 001, each prediction model 110Z calculated the values ​​of menopausal symptoms as 90%, sleep disorders as 21%, physical inactivity as 40%, dehydration as 76%, decreased immunity as 72%, and malnutrition as 11%.

[0540] As described above, by using the teaching data obtained by associating the purchase history of each customer, i.e., the purchase data 150Z, with the health status of these customers, the former is used as an explanatory variable and the latter is used as a target variable for machine learning, thereby constructing the prediction model 110Z. The method for constructing the prediction model 110Z will be described later in the following 3.

[0541] The output unit 130Z outputs the value calculated by each prediction model 110Z as a health prediction value 160Z related to the customer to the suggestion unit 135Z, a built-in storage medium, other information processing devices connected via a network, etc. Alternatively, the output unit 130Z may display the calculated health prediction value 160Z on a display device. Fig.15 In the example of , the output unit 140Z arranges the values ​​of customer ID, menopause, sleep disorder, etc. in a row direction, and repeats this arrangement for each ID of the customer data, thereby outputting health prediction values ​​160Z for multiple customers. In addition, the health prediction value 160Z includes not only the customer's current health status, but also the health status that can be expected to occur in the future (potential health status).

[0542] The suggestion unit 135Z suggests food, articles, services, etc. corresponding to the predicted health prediction value 160Z for each customer. Various suggestion methods are considered, such as displaying the suggestion on various terminals used by the store clerks, directly sending the suggestion content to the customer through a messenger service, or displaying the suggestion content to the customer on a web page.

[0543] 2.3 Specific examples and handling of various suggestions to customers

[0544] Below, refer to Figures 18 to 20 To illustrate the various suggestions that the comprehensive health prediction program 100Z makes to customers.

[0545] Fig.18 1 is a diagram schematically showing foods / articles / services etc. recommended by the recommendation unit 135Z according to the health prediction value 160Z of the customer. For example, for a woman (customer ID=001) whose menopausal symptoms are predicted to be 90%, the recommendation unit 135Z recommends products for improving menopausal symptoms sold in the drugstore 41Z. In addition, for a man (customer ID=002) whose nutritional deficiency is predicted to be 93%, the recommendation unit 135Z recommends high-function nutritional foods sold in the drugstore 41Z. As another example, for a woman (customer ID=003) whose predicted values ​​for menopause and sleep disorders are 79% and 89% respectively, the recommendation unit 135Z recommends visiting a medical institution 43Z. In addition, for customers with a relatively high degree of physical inactivity, the recommendation unit 135Z recommends going to the gym 45Z, and for customers with a degree of physical inactivity below a specified value, the recommendation unit 135Z recommends medical insurance products corresponding to lifestyle diseases provided by the insurance company 47Z.

[0546] Thus, the recommendation unit 135Z of the comprehensive health prediction program 100Z obtains prediction results from the plurality of prediction models 110Z respectively based on the purchase data 150Z, thereby being able to recommend food / articles / services suitable for the customer. In order to implement such a recommendation, a list of food, articles and / or services associated with each health symptom as an output value of the prediction model may be prepared in advance.

[0547] Fig.19 Table 50Z and databases 52Z, 54Z, and 56Z to be referred to for implementing advice related to menopausal symptoms are shown.

[0548] Table 50Z contains a menopause prediction model ( Fig.16 The output range of the health prediction model 110a for women in the example is 3 levels: level 50a with a value less than 40%, level 50b with a value greater than 40% and less than 80%, and level 50c with a value greater than 80%. In the case of level 50a, the drugstore's product database 52Z for preventing menopausal symptoms is referenced. The product database 52Z describes products X1 and X2 of product category P1 and product X3 of product category P2. The suggestion unit 135Z is able to refer to the product database 52Z and recommend some or all of the categories and / or product names to the corresponding customers. When the output value of the prediction model meets level 50b, the suggestion unit 135Z recommends products Y1 and Y2 of product category Q1 with reference to the product database 54Z for improving menopausal symptoms. When the output value of the prediction model meets level 50c, the suggestion unit 135Z uses the information of the current location and / or the residence of the pre-registered customer to recommend a medical institution with reference to the menopausal outpatient medical institution database 56Z. That is, this is equivalent to introducing the customer to a medical institution that performs preventive diagnosis and / or treatment according to the health prediction result of the customer.

[0549] Reference Fig. 20 Let me illustrate another example. Fig. 20 Table 60Z is shown for implementing advice corresponding to the output value of the lifestyle-related disease prediction model 110c.

[0550] Table 60Z is provided with the output value of the lifestyle disease prediction model 110c, in other words, with three levels of insufficient physical activity, namely, level 60a, which is less than 20%, level 60b, which is more than 20% and less than 80%, and level 60c, which is more than 80%. In the case of level 60a, it is considered that the physical activity is sufficient, and in order to continue the physical activity, nutrition guidance (1) and the monthly membership plan of the E1 gym are recommended. Nutrition guidance (1) is, for example, a recommendation for a diet with a nutrient component of less than 10 grams of lipids and more than 30 grams of protein per meal. In the case of level 60b, it is considered that the physical activity is not insufficient, and in order to carry out nutrition guidance (2) and regularly carry out physical activity, the membership plan of the E2 fitness club is recommended. Nutrition guidance (2) is, for example, a recommendation for a diet with a nutrient component of less than 20 grams of lipids and more than 20 grams of protein per meal. In the case of level 60c, based on the possibility of suffering from lifestyle diseases in the future due to insufficient physical activity, nutrition guidance (3) and medical insurance products proposed by E3 Life Insurance Company are recommended. Nutritional guidance (3) is, for example, a recommendation for a diet that contains less than 20 grams of lipids, more than 20 grams of protein, and less than 20 grams of carbohydrates per meal. Nutritional guidance, gyms, fitness clubs, and insurance product recommendations are the categories of products or services provided. Regarding the output value of the lifestyle disease prediction model 110c, the recommendation unit 135Z can also provide recommendations to the customer corresponding to the output value by referring to the table 60Z.

[0551] Apart from Fig.19 and Fig. 20 In addition to the example, similar tables and databases can be set up for each state, such as sleep state, immune state, water state, and nutritional state, so that the recommendation unit 135Z can provide recommendations to customers according to the health prediction value 160Z which is the output value of each prediction model 110Z for each customer.

[0552] In addition, the comprehensive health prediction program 100Z itself does not necessarily need to include the recommendation unit 135Z. For example, the following situation can also be considered: the output unit 130Z of the comprehensive health prediction program 100Z sends the health prediction value 160Z to a business operator different from the business operator using the comprehensive health prediction program 100Z based on the customer's permission, and the program of the different business operator installs the corresponding program in the recommendation unit 135Z to make the above-mentioned recommendations to the customer. For example, the comprehensive health prediction program 100Z can also send the health prediction value 160Z as the output value of the prediction model 110Z to an insurance company, so that the insurance company recommends insurance products to the customer.

[0553] 2.4 Processing Flow

[0554] Below, refer to Fig.21 The processing flow of the comprehensive health prediction program 100Z will be described. Fig.21 100 is a flowchart showing the procedure of the customer's health prediction process executed by the comprehensive health prediction program 100Z.

[0555] The input unit 120Z of the comprehensive health prediction program 100Z obtains purchase data 150Z (S701) indicating the purchase history of a customer who wants to make a prediction of the health status. For example, when purchasing goods at a retail store, the input unit 120Z reads the customer's membership card from the POS (Point of Sale) system, thereby constructing a database that associates the customer with the purchase history. With the customer's consent, the input unit 120Z can input the purchase data 150Z to the comprehensive health prediction program 100Z by extracting the purchase data 150Z from such a database. As described above, such purchase data 150Z includes, for example, JICFS classifications and commodity categories.

[0556] The commodity category determination unit 125Z determines the JICFS classification (classification code) and commodity category in the purchase data 150Z as inputs to each prediction model 110Z (S703). In addition, as described above, when the entire purchase data 150Z is used without specifically limiting the commodity category used for prediction by the prediction model 110Z, the processing of S703 is not required.

[0557] Next, each prediction model 110Z uses the classification code determined by the commodity category determination unit 125Z and the purchase data 150Z of the commodity category to predict the health status (S705). The output unit 130Z outputs the health prediction value 160Z indicating the health status predicted by the prediction model 110Z, and the suggestion unit 135Z provides the customer with suggestions based on the health prediction value 160Z (S707). Specifically, it is possible to consider the suggestions for improving health based on the health prediction value 160Z or the suggestion unit 135Z in the form of printing on a receipt, displaying on a display, notifying a smartphone, etc.

[0558] 2.5 Hardware Structure

[0559] Reference Fig. 22 A specific example of the hardware configuration of the information processing device 800Z that executes the comprehensive health prediction program 100Z will be described. Fig. 22 This is a hardware structure diagram of an information processing device 800Z that executes a comprehensive health prediction program 100Z. The information processing device 800Z uses the comprehensive health prediction program 100Z to calculate prediction results related to the customer's menopausal symptoms, sleep status, etc. The information processing device 800Z can be a commonly available computer system, such as a desktop PC (Personal Computer), a notebook PC, a tablet PC, a server computer, etc. The computer system can be set up in the office of an operator that uses the health prediction system 1Z, or it can be provided as a cloud service. In the latter case, a PC that can communicate with the cloud service can be set up in the office.

[0560] The information processing device 800Z includes a control unit 810Z, an input interface (I / F) unit 820Z, a storage device 830Z, and an output I / F unit 840Z.

[0561] The control unit 810Z can include a CPU (Central Processing Unit, not shown), a ROM (Read Only Memory, not shown), a RAM (Random Access Memory), etc. The control unit 810Z can execute the comprehensive health prediction program 100Z stored in the storage device 830Z. Thus, in addition to functioning as an ordinary computer, the information processing device 800Z can also perform various processes related to the above-mentioned health prediction. In addition, the calculation circuit included in the control unit 810Z can also be implemented by various processors such as MPU and GPU instead of CPU, and can also be implemented by multiple processors instead of one processor.

[0562] The input I / F unit 820Z receives the purchase data 150Z related to the customer who wants to predict the health status. As described above, the input I / F unit 820Z can receive the purchase data 150Z from an information processing device such as an external server connected via a network such as the Internet or a LAN. The input I / F unit 820Z can be implemented by, for example, an Ethernet (registered trademark) communication terminal, a USB (registered trademark) terminal, or a communication circuit that communicates in accordance with standards such as IEEE802.11, 4G, or 5G.

[0563] The storage device 830Z is a storage medium for storing computer programs and data required for the information processing device 800Z to operate. The storage device 830Z can be, for example, an HDD or an SSD which is a semiconductor storage device. The storage device 830Z can have a temporary storage element composed of a RAM such as a DRAM or SRAM, and can also function as an operating area of ​​the control unit 810Z. The storage device 830Z is used to store the comprehensive health prediction program 100Z and the data as a storage medium. Fig.19 Table 50Z and Fig. 20 A specific example of a recommended table 831Z is shown in Table 60Z.

[0564] The prediction model 110Z may be incorporated as a part of the comprehensive health prediction program 100Z, or may be provided as data separate from the comprehensive health prediction program 100Z. The prediction model 110Z may also be stored in the storage device 830Z in the form of a table or a parameter group.

[0565] The recommendation table 831Z may be a table that associates prediction results indicating the degree of health status, such as H level / M level / L level, 90% / 50% / 20%, with products / services to be recommended according to each degree.

[0566] The output I / F unit 840Z is a communication circuit and / or communication terminal connected to various output devices provided outside the information processing device 800Z. For example, the output I / F unit 840Z can be a USB terminal that outputs print data indicating the content to be printed by the receipt printer P. The comprehensive health prediction program 100Z can calculate the prediction result related to the health status of the customer, and send the result or the advice to the customer corresponding to the result to the receipt printer P to print on the receipt.

[0567] The output I / F unit 840Z may be a video output terminal connected to the display D. The comprehensive health prediction program 100Z calculates a prediction result related to the health status of the customer, and displays the result or a suggestion to the customer corresponding to the result on the display D. For example, by a pharmacist confirming the content displayed on the display D, the pharmacist can recommend one or more products corresponding to the symptoms of the customer.

[0568] As another embodiment, the output I / F unit 840Z may be a communication terminal or a communication circuit that is connected to a communication network N and capable of data communication. When the output I / F unit 840Z is a communication terminal or a communication circuit, the hardware of the input I / F unit 820Z and the output I / F unit 840Z may be the same. The output I / F unit 840Z may send the prediction result related to the health status of the customer or the advice to the customer corresponding to the result to the customer's smartphone M via a mobile phone line, for example.

[0569] 3 Learning prediction models 110Z learning system 9

[0570] 3.1 Overview

[0571] Next, refer to Fig.23 The method of generating the prediction model 110Z included in the comprehensive health prediction program 100Z by machine learning is described. Here, it is assumed that the model learning program 900Z generates the prediction model 110Z for illustration. In addition, the model learning program 900Z can be installed as a single program or as multiple programs that can operate collaboratively through the input and output of data.

[0572] also, Fig.15 The illustrated comprehensive health prediction program 100Z includes six prediction models 110Z, but these prediction models can be generated in the same manner using only different health data 954Z in the teaching data 950Z. The following description will focus on the case of generating a menopause prediction model 110a related to women's health.

[0573] 3.2 Teaching Data 950Z

[0574] The model learning program 900Z performs machine learning using teaching data 950Z including various data collected from a plurality of subjects to construct a prediction model 110Z. The teaching data 950Z includes purchase data 952Z related to the purchase behavior of each subject and health data 954Z which is data related to the health of the subjects.

[0575] Purchase data 952Z is ​​about and reference Fig.17The data of information similar to the information of the specific example is shown. That is, in the present embodiment, the items 150a related to the categories in the JICFS classification and the performance data 150b as numerical data such as the purchase amount, purchase quantity, and purchase frequency during the period can be included in the purchase data 952Z for each ID of the subject. In addition, since food and / or beverages are suitable as commodities associated with the health status of the subject or the customer, it is preferable to include the purchase history records of food and / or beverages in the purchase data 952Z.

[0576] Health data 954Z is used to describe the health status of the subject. The health status can be either a state related to subjective symptoms or a state in which the measured results are confirmed. That is, at least any data among health awareness, mental health, cognitive function, and health diagnosis results can be used as health data 954Z. In the health data 954Z of this embodiment, the judgment value of each state segment of the subject is described as the health status of the subject. In the state segment, each combination of the type and level (in other words, size) of the health status is defined. The type and level of the health status mentioned in this embodiment are the type and level of the physical and mental disorders of the corresponding individual. The judgment value can be shown as an index representing the health status using 2 levels or 3 levels, etc. In addition, the "type and level of health status" mentioned here does not have medical rigor, and can also be understood as the type and level of health troubles that an individual has.

[0577] Here, as an example, consider the case where the model learning program 900Z generates a menopause prediction model 110a and uses the symptoms of the simplified menopause index SMI to determine the level of menopause. In this case, the health data 954Z can include a label representing the symptoms of the simplified menopause index SMI to which the subject belongs. For example, the simplified menopause index SMI can be expressed as 2 levels. Specifically, "H" and "other than H" indicate relatively large menopausal symptoms. Regarding "other than H", it can also be subdivided into "M" with relatively moderate symptoms and "L" with relatively small symptoms (that is, the symptoms are expressed as a total of 3 levels). Below, the subject layer that meets "H" is referred to as the "H layer".

[0578] In addition, in the labels indicating the degree of symptoms in the health data 954Z, instead of 2 or 3 levels, the levels may be further subdivided into 4 or more levels, or numerical values ​​indicating the possibility of meeting menopausal symptoms may be considered to replace these levels.

[0579] That is, the teaching data 950Z is data obtained by associating the purchase data 952Z with the health data 954Z, wherein the purchase data 952Z includes information related to each commodity category of each subject, and information related to the purchase amount, purchase quantity, purchase frequency, etc. within a specified period, and the health data 954Z includes at least information indicating a label of the health status. Here, if the model learning program 900Z generates six prediction models 110Z of a woman's health status (menopausal symptoms), sleep status (sleep disorder), lifestyle disease (lack of physical activity), water status (degree of water deficiency), immune status (degree of immune decline), and nutritional status (degree of nutritional deficiency), for example, then the health data 954Z includes information corresponding to each of these prediction models 110Z. In machine learning in supervised learning, the teaching data 950Z is roughly classified into explanatory variables and target variables. In the present embodiment, the purchase data 952Z is ​​equivalent to the explanatory variable, and the health data 954Z is equivalent to the target variable. In addition, in the present embodiment, the explanatory variable only uses the purchase data 952Z, but is not limited to this. For example, consider adding gender and age as attributes of a subject to explanatory variables.

[0580] Next, the model learning program 900Z can generate a prediction model 110Z related to various symptoms such as sleep status, lifestyle-related diseases, immune status, water status, nutritional status, dementia, oxygen utilization, blood vessel health, hair health, intestinal environment, promotion of exercise effect, appetite improvement, eye health, skin health, oral hygiene, mental health, depression, stress, headache, shoulder stiffness, ear health, joint health, body temperature homeostasis, fatigue, bone health, male menopausal disorders, hypertension, weather pain, etc., as long as appropriate health data 954Z is prepared. In the following, indices that can be used as indicators of women's health status (menopausal symptoms), sleep status (sleep disorders), lifestyle-related diseases (lack of physical activity), water status (degree of water deficiency), immune status (degree of immune decline), and nutritional status (degree of nutritional deficiency) are specifically exemplified.

[0581] 3.2.1 Women’s health status (menopausal symptoms)

[0582] As an index indicating the degree of menopausal symptoms, for example, the simplified menopausal index SMI can be used as described above. The simplified menopausal index SMI is obtained by having each subject perform corresponding items corresponding to the degree of various symptoms listed in advance and scoring the results. The simplified menopausal index SMI is set as an index reflecting symptoms unique to Japanese menopausal women. Therefore, depending on the country, region, etc., it is also considered to appropriately use other indicators related to menopausal symptoms, such as the Menopause Rating Scale, the Kupperman index, the Green Climacteric Scale, the PSST (The premenstrual symptoms screening tool), MRS (Menopause rating Scale), WHQ (The Women's Health Questionnaire), VAS (Visual analogue scale), HFRDI (Hot Flash Related Daily Interference Scale), HFCS (Hot Flash Composite Score), MENQOL (Menopause-Specific Quality of life), and any one or more of the Japanese Women's Menopausal Symptom Evaluation Form.

[0583] In addition, regarding the health status of women, in addition to menopausal symptoms, premenstrual syndrome (PMS, PMDD) symptoms are also considered. That is, in order to predict premenstrual syndrome, the health prediction model 110Z can be constructed by performing machine learning using an index representing the degree of symptoms of premenstrual syndrome as a target variable and the purchase data 952 as an explanatory variable.

[0584] In addition, the health data 954Z may also include numerical data of these indicators, or may include only labels indicating the degree of symptoms according to the numerical values ​​of these indicators. This is also the case for indicators such as sleep status and lifestyle-related diseases in the future.

[0585] 3.2.2 Sleep status (sleep disorder)

[0586] As an indicator representing sleep, an indicator indicating the deterioration of the sleep state and / or the sleep rhythm, or any one of the inappropriate sleep time, is considered. It can be either the subjective symptoms of each control obtained through a questionnaire survey, etc., or a measured value confirmed by the results obtained by measuring each control. For the sleep state, at least one of the Athens Insomnia Scale (AIS), the Pittsburgh Sleep Quality Index (PSQI), the 3-Dimensional Sleep Scale (3DSS), and the Insomnia Severity Index (ISI) can be used. The degree of the symptoms is analyzed in the form of each item alone and each item in combination, and the health status of the consumer is classified by category. The Munich ChronoTypeQuestionnaire (μMCTQ) is used to find the central sleep time on weekdays and weekends, and the social jet lag (SJL) is calculated based on the difference between them, thereby making it possible to grasp the sleep rhythm.

[0587] 3.2.3 Lifestyle diseases (lack of physical activity)

[0588] As an index representing the exercise state, one or more indexes selected from lack of exercise, obesity, and changes in body weight over a certain period of time can be included. More specifically, it is considered to use one or more of the weight, BMI (Body Mass Index), number of steps, and / or metabolic syndrome diagnostic benchmark values ​​of each subject. The metabolic syndrome diagnostic benchmark value is one or more of the waist circumference diameter, the threshold value specified for diagnosing hypertriglyceridemia and / or low high-density lipoprotein cholesterol, the maximum and / or minimum blood pressure, and the threshold value specified for diagnosing fasting hyperglycemia.

[0589] Furthermore, when machine learning is used to generate a prediction model related to lifestyle diseases and exercise status, in addition to purchase data 952Z, information on one or more of whether or not the subject smokes or has seating behavior may be included in the explanatory variables for machine learning.

[0590] 3.2.4 Water status (degree of water deficiency)

[0591] As an index representing water status, it is considered to be serum sodium value, BUN / creatinine ratio and / or urine osmotic pressure, urine specific gravity, urine color, dehydration evaluation scale, and subjective symptoms caused by lack of water. These information can be obtained by the same method as the previous example of menopausal symptoms. For example, multiple controls are obtained by clinical examination values ​​obtained in health diagnosis, physical examinations, or health questionnaires implemented separately for a certain period, such as the physical changes experienced in a year. On the other hand, the subjective symptoms caused by the decline of water can be obtained by questionnaires, etc. The degree of symptoms can be grasped according to at least one of serum sodium value, urine osmotic pressure, urine specific gravity, BUN / creatinine ratio, water intake, physical activity, and urine color.

[0592] Based on these indicators, it is possible to find out the health status that has deteriorated or is likely to be deteriorated in the future through epidemiological information from literature. For example, it is found that consumers with serum sodium values ​​exceeding 142mEq / L have a higher risk of cognitive impairment and hypertension than consumers with serum sodium values ​​less than 142mEq / L, and can be evaluated as having a high health risk.

[0593] In addition, when generating machine learning of a predictive model related to water status, in addition to purchase data 952Z, machine learning can also be performed by including one or more information of the subject's water intake, alcohol intake, physical activity, urine color, urine osmotic pressure, urine specific gravity, serum sodium value, and BUN / creatinine ratio in the explanatory variables.

[0594] 3.2.5 Immune status (degree of immune decline)

[0595] As an indicator of immune status, for example, cold susceptibility and / or its frequency, cold symptoms can be used. The cold susceptibility and frequency can be obtained from the subjective symptoms of each control obtained through a questionnaire survey, etc., or from the medical institution's medical records. The degree of symptoms is statistically divided into more than three stages, and the degree of symptoms can be cold susceptibility, SIgA concentration, allergic symptoms, oral environment and immune status, exercise status, stress status, severe sleep symptoms, moderate sleep symptoms, mild sleep symptoms, and cold symptoms.

[0596] 3.2.6 Nutritional status (degree of malnutrition)

[0597] The index indicating the nutritional status may be the subjective symptoms of each control obtained through a questionnaire survey or the like, or may be a measured value confirmed by measuring the results of each control. The degree of the symptom may be at least one of the presence or absence of subjective symptoms caused by decreased nutrition, the diversity score of food intake (DVS), the shortened version of the appetite questionnaire (SNAQ: Simplified Nutritional Appetite Questionnaire), and the BMI (Body Mass Index).

[0598] 3.3 Machine Learning Methods

[0599] The model learning program 900Z involved in this embodiment uses logistic regression as a method of machine learning. However, logistic regression is an example, and other methods are also considered, such as random forest, decision tree, gradient boosting, support vector regression, linear regression, partial least squares (PLS) regression, Gaussian process regression, neural network, etc. In addition, when using logistic regression, the accuracy can be further improved by using L1 normalization, and the logistic regression (ridge regression) using L2 normalization instead of L1 normalization can also be sampled. It is also possible to use an elastic net that uses both L1 normalization and L2 normalization.

[0600] Logistic regression is a method that accepts input and determines whether it corresponds to an event by calculating the probability that it corresponds to a certain event (or the probability that it does not correspond to it). As an example, the model of logistic regression is described as the following equation (1) called a logistic function. In equation (1), the left side is a natural logarithm. In addition, i represents the i-th subject, p represents the probability of the event of the target variable, b1, b2, ..., b n represents the partial regression coefficient, b0 represents the constant term, x1, x2, ···, x n Denotes a descriptive variable.

[0601] [Number 1]

[0602]

[0603] If the right side of equation (1) is set to z, equation (1) can be transformed into equation (2) representing probability p.

[0604] [Number 2]

[0605]

[0606] Formula (2) is called the sigmoid function, which is the activation function in the logistic regression model. The partial regression coefficients b1, b2, . . . , b in the definition of z aren is called a weight, and b0 is a bias term. If z is input to the sigmoid function, a value between 0 and 1, i.e., a probability value, is calculated.

[0607] Here, the S-shaped function obtained by equation (2) is replaced by According to its output value, it is classified into output y representing any value of two values. Formula (3) means that if If it is 0.5 or more, it is classified as level 1. If it is less than 0.5, it is classified as level 0. In addition, 0.5 is an example, and a value other than 0.5 can also be set as the threshold. In other words, formula (3) means that if z in formula (2) is greater than 0, it is classified as level 1, and if z in formula (2) is less than 0, it is classified as level 0.

[0608] [Number 3]

[0609]

[0610] Next, we introduce the "likelihood" used for learning in logistic regression. Likelihood refers to the likelihood of a condition when looking at the result. The likelihood function L representing the likelihood is represented by the following equation (4). In addition, P represents a probability value.

[0611] [Number 4]

[0612]

[0613] The likelihood function L represents the probability of accurately judging all events. By finding the weight that maximizes the likelihood, the probability of the event you want to predict can be output more accurately. Specifically, the likelihood function L is multiplied by (-1) to reverse the sign of the likelihood function. The likelihood function obtained by reversing the sign is the error function in logistic regression. In order to find the weights that minimize the error function, that is, the partial regression coefficients b1, b2, ..., b n , in the error function for b1, b2···b n Partial differential and gradient descent methods are applied respectively. Thus, logistic regression learning is performed.

[0614] Alternatively, as the likelihood function, a log-likelihood function using the natural logarithm of equation (4) may be used. The optimal weight can be found by minimizing the function obtained by multiplying the log-likelihood function by (-1) and reversing the sign.

[0615] The calculation algorithm related to the above-mentioned logistic regression is well known, and software applications for performing machine learning using such algorithms can be easily obtained. In principle, it is as described above, but by using available software, even a person who does not know the details of the specific analytical method using the formula can realize the machine learning using logistic regression involved in this embodiment.

[0616] 3.4 Method for determining product categories for health prediction

[0617] The numerical data such as the purchase amount, purchase quantity, and purchase frequency related to each commodity and service category in the purchase data 952Z included in the teaching data 950Z corresponds to the explanatory variable x in the above-mentioned formula (1) and the like.

[0618] As described above, regarding the health data 954Z included in the teaching data 950Z of the present embodiment, for each symptom such as menopausal symptoms and sleep disorders, each subject is assigned a label of "H", "L" (or "H", "M", "L") indicating whether the subject is classified as the H layer or the L layer. In the above formula (3), the likelihood function can be obtained by classifying in a manner such that "1" is output in the case of the H layer and "0" is output in the case of a layer other than the H layer. The elements constituting the explanatory variable x are not limited to the above-mentioned purchase data 952Z. For example, it is also considered that the explanatory variable includes data indicating age group and data indicating gender.

[0619] By performing machine learning using the teaching data 950Z related to a plurality of subjects to construct a prediction model, it is possible to identify a group (category) of products that are likely to be purchased by customers in the H layer.

[0620] Fig.24 The following table shows the product categories selected (most likely to be purchased) by most subjects in the H layer and the layers other than the H layer related to menopausal symptoms. Fig.24 The upper part shows the commodity categories selected by the majority of the subjects in the H layer, and the lower part shows the commodity categories also selected by the subjects in the layers other than the H layer.

[0621] in addition, Fig.25 The relationship between menopausal symptoms and health issues found from the purchased items of subjects in the H layer related to menopausal symptoms is shown. Fig.25The numerical value in the far right column shows the degree of ease (bias value) of being selected by women with menopausal symptoms. This numerical value is quoted from "Symptom-specific bias values ​​for items that are dealt with even if it costs money - Report on the research project on women's understanding of health and beauty lifestyles - Behavior observation survey results (Symptom bias values ​​for different items that people want to deal with even if it costs money - Research project report on women's understanding of health and beauty lifestyles - Behavior observation survey results)".

[0622] These results show that the subjects in the H group who were concerned about menopausal symptoms actively selected (purchased) specific products. The subjects in the H group purchased products corresponding to the self-reported symptoms of dryness and irregularity of the skin or body.

[0623] Fig.26 An example of a product group (product category) for women with menopausal symptoms belonging to the H layer is shown. In this way, it is possible to determine the product category that subjects in the H layer frequently purchase compared to subjects in layers other than the H layer, and therefore it can be inferred that subjects who purchase more (quantity, amount, number of times) of such products are likely to belong to the H layer of menopausal symptoms. In addition, similarly, it is also possible to determine the product category that subjects who do not belong to the H layer frequently purchase compared to subjects in the H layer, and therefore it can be inferred that subjects who purchase more of such product categories are likely not to belong to the H layer of menopausal symptoms.

[0624] Therefore, the model learning program 900Z in this embodiment constructs a prediction model 110Z using data in the purchase data 952Z that can help infer whether a product category belongs to the H layer or not, and can infer customers belonging to the H layer with high accuracy.

[0625] 3.5 Structure of the Model Learning Program 900Z

[0626] Below, refer to Fig.23 The functional structure of the model learning program 900Z will be explained. Fig.23 900Z. The model learning program 900Z includes an input unit 910Z, a commodity category identification unit 912Z, a learning unit 914Z, and an output unit 916Z. The model learning program 900Z may be implemented as a single program or as a plurality of programs that cooperate with each other.

[0627] The input unit 910Z receives input of teaching data 950Z. The input unit 910Z may read the teaching data 950Z from a storage medium such as a HDD or SSD built into an information processing device (computer) that executes the model learning program 900Z, or may receive input of teaching data 950Z from an external information processing device connected via a network such as a LAN or the Internet.

[0628] The commodity category determination unit 912Z selects the commodity category of the purchase data 952Z used in the machine learning of the prediction model 110Z as the learning object. The method of selecting the commodity category is described in the above 3.4, so the detailed description is omitted here. In addition, when a plurality of prediction models 110Z corresponding to a plurality of symptoms are generated, the commodity category determination unit 912Z can select a different commodity category for each prediction model 110Z.

[0629] As described above, when the prediction model 110Z is machine-learned, it is also possible to create the prediction model 110Z using the entire purchase data 952Z without limiting the commodity category. In this case, the commodity category identification unit 912Z is ​​not required.

[0630] The learning unit 914Z performs machine learning using the purchase data 952Z of the commodity category identified by the commodity category identification unit 912Z to generate a prediction model 110Z. The method of generating the prediction model 110Z is described in 3.3 above, and therefore a detailed description thereof is omitted here.

[0631] The output unit 916Z outputs the prediction model 110Z generated by the learning unit 914Z to another information processing device (for example, see Fig. 22 The information processing device 800Z described above) etc. In addition, as an output method, the case where the output is in the form of the function itself constituting the prediction model 110Z, or the case where the output is in the form of the parameters of the function constituting the prediction model 110Z is also considered.

[0632] 3.6 Processing Flow

[0633] Below, refer to Fig. 27 The processing flow of the model learning program 900Z will be described. Fig. 27 is a flowchart showing the process involved in creating a prediction model 110Z performed by the model learning program 900Z.

[0634] The input unit 910Z of the model learning program 900Z acquires teaching data 950Z related to a plurality of subjects for learning (S1301). The teaching data 950Z includes purchase data 952Z indicating the purchase history of the subject and health data 954Z indicating the health status (degree of symptom). As described above, the purchase data 952Z may include, for example, JICFS classification (classification code) and commodity category, and numerical data such as the number of purchases, purchase amount, and purchase quantity corresponding thereto.

[0635] Next, the commodity category determination unit 912Z determines the JICFS classification and commodity category used in the generation of the prediction model 110Z (S1303). The determination method is as described in 3.4 above, but the commodity category determination unit 912Z only needs to select the JICFS classification and commodity category that have a great influence on the inference of the health status obtained as the health data 954Z. In addition, when the prediction model 110Z is created using the purchase data 952Z as a whole without specifically limiting the commodity category, the processing of S1303 is not required.

[0636] The learning unit 914Z performs machine learning using the purchase data 952Z as explanatory variables and the health data 954Z as the target function (S1305). The machine learning method has been described with a specific example in 3.3 above, so the description is omitted here.

[0637] The output unit 916Z outputs the prediction model 110Z generated by the learning unit 914Z in S1305 to a built-in storage medium, another information processing device connected via a network, etc. In addition, as described above, the prediction model 110Z can be output in the form of the function itself constituting the prediction model 110Z, or in the form of parameters, etc.

[0638] 3.7 Hardware Structure

[0639] Reference Fig.28 A specific example of the hardware configuration of the information processing device 1400Z that executes the model learning program 900Z will be described. Fig.28 This is a hardware configuration diagram of an information processing device 1400Z that executes a model learning program 900Z. The information processing device 1400Z generates a prediction model 110Z that predicts a customer's menopausal symptoms, sleep state, and the like using the model learning program 900Z.

[0640] The information processing device 1400Z is the same as the above-mentioned information processing device 800Z in the following aspects, etc.: it can be implemented as a generally available computer system; it can be installed in the office of the operator who uses the health prediction system 1Z, and it can also be provided as a cloud service. In addition, in this embodiment, the information processing device 800Z for performing health prediction and the information processing device 1400Z for generating the prediction model 110Z are different devices, but the two can also be implemented as the same device.

[0641] The information processing device 1400Z includes a control unit 1410Z, a communication I / F unit 1420Z, and a storage device 1430Z.

[0642] The control unit 1410Z can include a CPU, a ROM, a RAM, etc. The control unit 1410Z can execute the model learning program 900Z stored in the storage device 1430Z. Thus, in addition to functioning as a normal computer, the information processing device 1400Z can also perform various processes involved in the construction of the above-mentioned prediction model 110Z. In addition, the calculation circuit included in the control unit 1410Z can also be implemented by various processors such as MPU and GPU instead of CPU. In addition, it can also be implemented by multiple processors instead of one processor.

[0643] The communication I / F unit 1420Z and the input I / F unit 820Z can be implemented by, for example, an Ethernet (registered trademark) communication terminal, a USB (registered trademark) terminal, or a communication circuit that communicates in accordance with standards such as IEEE802.11, 4G, or 5G. The communication I / F unit 1420Z can be connected to a communication network such as an intranet or the Internet to receive teaching data 950Z prepared by an operator of the operation learning system 9. In addition, the information processing device 1400Z can also communicate directly with other devices via the communication I / F unit 1420Z, or can communicate via an access point or the like.

[0644] The storage device 1430Z is a storage medium for storing computer programs and data required for the information processing device 1400Z to operate. The storage device 1430Z can be, for example, an HDD or an SSD. The storage device 1430Z can have a temporary storage element composed of a RAM such as a DRAM or an SRAM, and can also function as an operating area of ​​the control unit 1410Z. The storage device 1430Z is used to store the model learning program 900Z and the prediction model 110Z generated by the model learning program 900Z (including a table and / or a parameter group for defining the prediction model 110Z).

[0645] 4 Effects of this embodiment

[0646] The health prediction system 1Z according to this embodiment can predict multiple health risks of a customer based on the customer's purchase data 150Z in the comprehensive health prediction program 100Z, and can provide detailed advice to each customer for improving and / or maintaining the current or future health status based on each prediction result.

[0647] 5. Notes

[0648] Furthermore, the configurations of the above-described embodiments may be combined or some of the components may be replaced. In addition, the configuration of the present invention is not limited to the above-described embodiments, and various modifications may be made within the scope of the gist of the present invention.

[0649] For example, in the above-mentioned embodiment, it is assumed that the comprehensive health prediction program 100Z includes 6 health prediction models, but the case of including 6 is just an example. It may also include more than 2 and less than 6, or more than 7 health predictions. If generally recorded, the comprehensive health prediction program 100Z can include prediction models 110Z related to each health symptom of the 1st to Nth (N is an integer greater than 2). Moreover, the health prediction model 110Z related to the kth (1≤k≤N) health symptom is constructed as a prediction model that takes the purchase data 150Z related to the commodity category to which at least one commodity belongs as input and the degree of the kth health symptom as output. The health prediction model 110Z related to the kth (1≤k≤N) health symptom can be constructed by using the purchase data 952Z related to the purchase history records of each of the multiple subjects as an explanatory variable and the health data 954Z representing the degree of health status collected from these subjects as the target variable for machine learning.

[0650] In addition, when the degree of health symptoms is expressed by numerical values, the upper and lower limits of the numerical range for defining the level are examples and can be changed as appropriate. The number of defined levels can also be arbitrarily determined by those skilled in the art.

[0651] As can be easily understood by those skilled in the art having the benefit of the present disclosure, the above-mentioned various examples can be appropriately combined with each other in various modes as long as no contradiction occurs.

[0652] It is to be understood that, if many possible embodiments to which the principles of the invention disclosed in this specification can be applied are considered, the various embodiments illustrated are merely preferred examples, and the scope of protection of the invention involved in the claims should not be considered to be limited to these preferred examples. In fact, the scope of protection of the invention involved in the claims is determined by the attached claims. Therefore, all contents belonging to the scope of protection of the invention recorded in the claims are requested to be authorized as the invention of the present inventors.

[0653] The present application is based upon and claims the benefit of priority from the following Japanese patent applications, the entire contents of which are incorporated herein by reference.

[0654] (1) Japanese Patent Application No. 2022-046894, filed on March 23, 2022, entitled “Computer Program, Information Processing Device, and Method”

[0655] (2) Japanese Patent Application No. 2022-046895, entitled “Program, Information Processing Device, and Information Processing Method,” filed on March 23, 2022

[0656] Description of Reference Numerals

[0657] 1: Communication system; 2: Communication network (communication line); 10, 10A, 10B: Server device; 20, 20A, 20B: Terminal device; 11, 21: Central processing unit (CPU).

Claims

1. A computer program, characterized in that By being executed by at least one processor, the at least one processor performs the following functions: acquiring subject behavior data for identifying a history of behaviors performed by a subject user; and By inputting the subject behavior data into an inference model generated by performing supervised learning, status data for identifying the water replenishment status of the subject user is output from the inference model, the status data including at least one of data for identifying the serum sodium value of the subject user, data for identifying the urine osmotic pressure of the subject user, data for identifying the urine specific gravity of the subject user, data for identifying the urine color of the subject user, data for identifying the urea nitrogen (BUN / creatinine ratio) of the subject user, data for identifying a dehydration evaluation scale of the subject user, and data for identifying subjective symptoms associated with insufficient water of the subject user.

2. A computer program, characterized in that By being executed by at least one processor, the at least one processor performs the following functions: acquiring subject behavior data for identifying a history of behaviors performed by a subject user; and By inputting the object behavior data into an inference model generated by performing supervised learning, status data for identifying the immune status of the object user is output from the inference model, the status data including data for identifying the object user's susceptibility to colds, data for identifying the frequency of colds of the object user, data for identifying the SIgA concentration of the object user, data for identifying the allergic symptoms of the object user, data for identifying the status of the object user's oral environment and / or immunity, data for identifying the object user's sleep status, data for identifying the object user's movement status, and at least one of data for identifying the object user's stress status.

3. A computer program, characterized in that By being executed by at least one processor, the at least one processor performs the following functions: acquiring subject behavior data for identifying a history of behaviors performed by a subject user; and By inputting the object behavior data into an inference model generated by performing supervised learning, status data for identifying the nutritional status of the object user is output from the inference model, the status data including data for identifying the presence or absence of subjective symptoms caused by nutritional decline of the object user, data for identifying the diversity score (DVS) of the object user's food intake, data for identifying the shortened version of the appetite scale (Simplified Nutrition Appetite Questionnaire) of the object user, and at least one of data for identifying the body mass index (BMI) of the object user.

Images