Measure proposal device, measure proposal method, measure proposal program, creation method, and learned model

Abstract

The present invention addresses the problem of providing a measure proposal device or the like capable of proposing a personalized behavior modification measure for improving quality of life (in particular, dietary habits, physical activity, or sleep). A measure proposal device according to the present embodiment acquires actual dietary habit data relating to the actual dietary habits of a user, acquires intervention measure data relating to an intervention measure by the user for a health matter desired by the user, acquires a diet evaluation score relating to the quality of the current diet of the user using a machine learning model on the basis of the acquired actual dietary habit data, generates intervention measure data relating to an intervention measure that can be added to the intervention measure on the basis of the acquired intervention measure data and the acquired diet evaluation score, and outputs the generated intervention measure data.

Description

Policy proposal device, policy proposal method, policy proposal program, creation method, and trained model

[0001] The present invention relates to a policy proposal device, a policy proposal method, a policy proposal program, a creation method, and a trained model.

[0002] Patent Document 1 discloses a behavioral change promotion device that can promote appropriate behavioral change for each user.

[0003] International Publication No. 2023 / 013475

[0004] However, Patent Document 1 cannot propose personalized behavioral change measures that improve quality of life.

[0005] The present invention has been made in consideration of the above-mentioned problems, and aims to provide a policy proposal device, a policy proposal method, and a policy proposal program that can propose personalized behavioral change measures that improve the quality of life (especially diet, physical activity, or sleep), as well as a creation method and trained model that can create a machine learning model that predicts an evaluation score regarding a user's current quality of life, which contributes to making such proposals.

[0006] In order to solve the above-mentioned problems and achieve the objectives, the policy proposal device of the present invention comprises a first acquisition unit that acquires lifestyle data related to a user's lifestyle, an evaluation unit that uses a machine learning model to acquire an evaluation score related to the user's current quality of lifestyle based on the lifestyle data acquired by the first acquisition unit, a generation unit that generates intervention policy data related to intervention policies based on the evaluation score acquired by the evaluation unit, and an output processing unit that outputs the intervention policy data generated by the generation unit.

[0007] The lifestyle data may be data relating to the user's dietary habits, and the evaluation score may be a diet evaluation score relating to the quality of meals.

[0008] The dietary evaluation score is a score for determining whether a diet is good or bad, and may be a score calculated from a plurality of nutrients or foods according to the health phenomenon desired by the user.

[0009] In addition, the dietary evaluation score may be the sum of a first evaluation score based on the content in the diet of a first nutrient that is preferable to be consumed in large amounts, a second evaluation score based on the content in the diet of a first food ingredient that is preferable to be consumed in large amounts, and a third evaluation score based on the content in the diet of a second nutrient that is preferable to be consumed in small amounts.

[0010] The first nutrient may be protein, the first food ingredient may be a vegetable, and the second nutrient may be saturated fatty acid and an amount of salt equivalent or sodium.

[0011] The health event may also be an eating behavior that the user wants to achieve.

[0012] The lifestyle data may relate to the user's actual physical activity, and the evaluation score may be an activity evaluation score relating to the quality of the physical activity.

[0013] The activity evaluation score is a score for determining whether the physical activity is good or bad, and may be a score calculated from a plurality of physical activity-related indices according to the health events desired by the user.

[0014] The activity evaluation score may also be the sum of a first evaluation score based on the ratio of actual exercise time to recommended exercise time, a second evaluation score based on exercise intensity, and a third evaluation score based on exercise frequency.

[0015] The health event may also be a physical activity that the user wishes to accomplish.

[0016] The lifestyle data may be data relating to the user's sleep habits, and the evaluation score may be a sleep evaluation score relating to the quality of sleep.

[0017] The sleep evaluation score is a score for determining whether the user's sleep is good or bad, and may be a score calculated from a plurality of sleep-related indices according to health events desired by the user.

[0018] The sleep evaluation score may also be the sum of a first evaluation score based on the ratio of actual sleep time to recommended sleep time, a second evaluation score based on the ratio of deep sleep, and a third evaluation score based on sleep interruptions.

[0019] The health event may also be a sleep goal that the user wishes to achieve.

[0020] In addition, the measure proposal device of the present invention may further include a second acquisition unit that acquires intervention measure data regarding the user's first intervention measure for the health event desired by the user, and the generation unit may generate intervention measure data regarding a second intervention measure that can be added to the first intervention measure based on the intervention measure data acquired by the second acquisition unit and the evaluation score acquired by the evaluation unit.

[0021] In addition, the policy proposal device of the present invention may further include a calculation unit that calculates the amount of change in the user's evaluation score due to the second intervention policy based on the intervention policy data acquired by the second acquisition unit and the evaluation score acquired by the evaluation unit, and the output processing unit may further output the amount of change calculated by the calculation unit.

[0022] In addition, the calculation unit may calculate the change in the first evaluation score, the change in the second evaluation score, and the change in the third evaluation score, and calculate the sum of the calculated changes as the change in the evaluation score.

[0023] The first acquisition unit may acquire the actual living situation data using a large-scale language model, and the second acquisition unit may acquire the intervention measure data using a large-scale language model.

[0024] The intervention measure data generated by the generation unit may be text data.

[0025] In addition, the policy proposal method of the present invention includes a first acquisition step in which a first acquisition unit acquires lifestyle data related to the user's lifestyle; an evaluation step in which an evaluation unit uses a machine learning model to acquire an evaluation score related to the user's current quality of lifestyle based on the lifestyle data acquired in the first acquisition step; a generation step in which a generation unit generates intervention policy data related to an intervention policy based on the evaluation score acquired in the evaluation step; and an output processing step in which an output processing unit outputs the intervention policy data generated in the generation step.

[0026] In addition, the policy proposal program of the present invention causes a computer to function as a first acquisition means for acquiring lifestyle data related to a user's lifestyle, an evaluation means for using a machine learning model to acquire an evaluation score related to the user's current quality of lifestyle based on the lifestyle data acquired by the first acquisition means, a generation means for generating intervention policy data related to intervention policies based on the evaluation score acquired by the evaluation means, and an output processing means for outputting the intervention policy data generated by the generation means.

[0027] Furthermore, the recording medium according to the present invention is a non-transitory computer-readable recording medium having the policy proposal program recorded thereon. In other words, the recording medium according to the present invention is a non-transitory computer-readable recording medium including programmed instructions for causing a computer to execute the policy proposal method.

[0028] In addition, the creation method of the present invention is characterized in that it creates a trained model by using learning data that associates lifestyle data related to a user's lifestyle with an evaluation score related to the quality of the user's current lifestyle, and executing a learning process for the parameters of a machine learning model that outputs an evaluation score when lifestyle data is input.

[0029] In addition, the trained model of the present invention is a trained model for outputting an evaluation score regarding the quality of a user's current living situation when actual living situation data regarding the user's living situation is input, and the parameters of the trained model are trained using learning data that associates actual living situation data with evaluation scores, and is characterized in that the trained model is configured to accept actual living situation data of a certain user as input, perform calculations based on the parameters for the input actual living situation data, and output an evaluation score.

[0030] In addition, the policy proposal device of the present invention comprises a first acquisition unit that acquires lifestyle data related to a user's lifestyle, an evaluation unit that uses a machine learning model to acquire an evaluation score related to the user's current quality of lifestyle based on the lifestyle data acquired by the first acquisition unit, a generation unit that generates intervention policy data related to intervention policies based on the evaluation score acquired by the evaluation unit, and an output processing unit that outputs the intervention policy data generated by the generation unit, wherein the machine learning model is a trained model created by the creation method.

[0031] The present invention has the effect of being able to propose personalized behavioral change measures that improve the quality of life (especially diet, physical activity, or sleep).Furthermore, the present invention has the effect of being able to create a machine learning model that predicts a user's current quality of life evaluation score, which contributes to the proposal, and to provide a trained model that contributes to the proposal.

[0032] FIG. 1 is a diagram illustrating an overview of this embodiment. FIG. 2 is a block diagram illustrating the configuration of the measure proposal device 100. FIG. 3 is a diagram illustrating processing executed by the measure proposal device 100. FIG. 4 is a diagram illustrating an example of content information data stored in the storage unit 106. FIG. 5 is a diagram illustrating an example of user information data stored in the storage unit 106. FIG. 6 is a diagram illustrating an example of a method for acquiring a diet evaluation score related to diet quality. FIG. 7 is a diagram illustrating an example of a method for acquiring an intervention content, a framework, a range of effects expected from the intervention content, an effect size of the effects expected from the intervention content, and an effect of an intervention measure on a diet evaluation score. FIG. 8 is a diagram illustrating an example of content information data acquired by the first acquisition unit 102a and intervention measure category data acquired by the second acquisition unit 102b. FIG. 9 is a diagram illustrating an example of user information data acquired by the first acquisition unit 102a and diet quality evaluation score data acquired by the evaluation unit 102c. Fig. 10 is a diagram showing an example of additional intervention measure data generated by the generation / calculation executing unit 102d and output by the output processing unit 102e, and example data on the amount of change in the predicted evaluation score of diet quality calculated by the generation / calculation executing unit 102d and output by the output processing unit 102e. Fig. 11 is a flowchart showing an example of a measure proposal process executed by the generation / calculation executing unit 102d. Fig. 12 is a diagram showing a first example of the measure proposal process. Fig. 13 is a diagram showing a second example of the measure proposal process. Fig. 14 is a diagram showing an example of similarity evaluation between texts.

[0033] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of a policy proposal device, a policy proposal method, and a policy proposal program according to the present invention will be described in detail with reference to the accompanying drawings. However, the present invention is not limited to the embodiments.

[0034] [1. Overview] FIG. 1 is a diagram showing an overview of this embodiment.

[0035] In this embodiment, as shown in FIG. 1 , the system interprets the behavioral change measures (nudges) that the user is already taking and obtains a diet quality evaluation score based on the user's data and input text, generates additional behavioral change measures recommended for improving diet quality, and calculates the change in the diet quality evaluation score resulting from the additional implementation of the behavioral change measures. This allows the system to propose personalized behavioral change measures for improving diet quality. While this embodiment describes how to propose behavioral change measures for improving diet quality when lifestyle is viewed from the perspective of diet, it may also propose behavioral change measures for improving the quality of physical activity (exercise or daily activities) or sleep when lifestyle is viewed from the perspective of physical activity or sleep.

[0036] 2. Configuration and Processing FIG. 2 is a block diagram showing the configuration of the policy proposal device 100. As shown in FIG.

[0037] The policy proposal device 100 is communicatively connected to information and communication terminals (e.g., mobile phones, smartphones, tablet terminals, personal computers, etc.) owned by users (ordinary people) via a network such as the Internet, an intranet, or a LAN (including both wired and wireless).

[0038] The policy proposal device 100 includes (1) a control unit 102 that comprehensively controls devices such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit), (2) a communication interface unit 104 that communicatively connects the device to a network via a communication device such as a router and a wired or wireless communication line such as a dedicated line, (3) a storage unit 106 that stores various databases, tables, files, etc., and (4) an input / output interface unit 108 that connects to an input device 112 and an output device 114. The units included in the proposal device 100 are communicatively connected via any communication path.

[0039] The control unit 102 has an internal memory for storing control programs such as an OS, programs that define various processing procedures, required data, etc., and executes various information processing based on these programs. Functionally, the control unit 102 conceptually includes a first acquisition unit 102a (corresponding to the first acquisition unit of the present invention), a second acquisition unit 102b (corresponding to the second acquisition unit of the present invention), an evaluation unit 102c (corresponding to the evaluation unit of the present invention), a generation / calculation execution unit 102d (corresponding to the generation unit and calculation unit of the present invention), and an output processing unit 102e (corresponding to the output processing unit of the present invention). Each processing unit included in the control unit 102 will be described later.

[0040] The communication interface unit 104 mediates communication between the policy proposal device 100 and a network (or a communication device such as a router). That is, the communication interface unit 104 has a function of communicating data with other terminals via a communication line.

[0041] The memory unit 106 is a storage means. For example, a memory device such as a RAM or ROM, a fixed disk device such as a hard disk, a flexible disk, or an optical disk can be used as the memory unit 106. The memory unit 106 may store a computer program that works in cooperation with an operating system (OS) to issue commands to a CPU or GPU to perform various processes. The memory unit 106 stores user information data and content information data. Each piece of data stored in the memory unit 106 will be described later.

[0042] An input device 112 and an output device 114 are connected to the input / output interface unit 108. The output device 114 may be a monitor, a speaker, a printer, or the like. The input device 112 may be a keyboard, a mouse, a microphone, a monitor that cooperates with a mouse to realize a pointing device function, a touch panel, or the like.

[0043] FIG. 3 is a diagram showing the process executed by the policy proposal device 100.

[0044] User information data and content information data are stored in the storage unit 106. The content information data stored in the storage unit 106 will be described in detail later with reference to Fig. 4. The user information data stored in the storage unit 106 will be described in detail later with reference to Fig. 5.

[0045] For examples of methods for assessing the quality of research using content information data, see Non-Patent Documents 1 and 2. For literature classification algorithms using content information data, see Non-Patent Document 3. - Non-Patent Document 1: Nitin Shivappa, Susan E Steck, Thomas G Hurley, James R Hussey and James R Hebert, "Designing and developing a literature-derived, population-based dietary inflammatory index", Public Health Nutrition: 17(8), 1689-1696, 14 August 2013 - Non-Patent Document 2: "The Maryland Scientific Methods Scale (SMS)", https: / / whatworksgrowth.org / resource-library / the-maryland-scientific-methods-scale-sms / - Non-Patent Document 3: "Welcome to the Theory & Techniques Tool", https: / / theoryandtechniquetool.humanbehaviourchange.org /

[0046] The first acquisition unit 102a acquires user information data and content information data. Specific details regarding the content information data acquired by the first acquisition unit 102a will be described later with reference to Fig. 8. Specific details regarding the user information data acquired by the first acquisition unit 102a will be described later with reference to Fig. 9.

[0047] For an example of a framework for acquiring desired data regarding the memory unit 106 and the first acquisition unit 102a, see Non-Patent Document 4. Non-Patent Document 4: Zhiyu Chen, Yujie Lu, and William Yang Wang, "Empowering Psychotherapy with Large Language Models: Cognitive Distortion Detection through Diagnosis of Thought Prompting," Findings of the Association for Computational Linguistics: EMNLP 2023, pages 4295-4304, December 6-10, 2023

[0048] The second acquiring unit 102b acquires intervention measure category data. The intervention measure category data acquired by the second acquiring unit 102b will be described in detail below with reference to FIGS.

[0049] The evaluation unit 102c acquires a diet evaluation score related to the quality of the user's current diet using a machine learning model based on the user information data (specifically, the actual dietary habits) acquired by the first acquisition unit 102a. The machine learning model may be, for example, a logistic regression model, a random forest model, or a generative AI model. The diet evaluation score is a score that determines the quality of a diet and may be a score calculated from multiple types (e.g., two, three, four, five, six, seven, eight, nine, or ten types) of nutrients or foods according to the health phenomenon desired by the user (e.g., the eating behavior the user wants to achieve). For example, the dietary evaluation score may be calculated from foods classified as "meat," "seafood," "eggs," "milk or dairy products," "oils and fats," "soybeans or soybean products," "green and yellow vegetables," "potatoes," "seaweed," and "fruit" (see https: / / balance-check.ajinomoto.co.jp / ). The dietary evaluation score may be the sum of a first evaluation score based on the dietary content of a first nutrient (e.g., protein) for which a higher intake is desirable, a second evaluation score based on the dietary content of a first food ingredient (e.g., vegetables) for which a higher intake is desirable, and a third evaluation score based on the dietary content of a second nutrient (e.g., saturated fatty acids, salt equivalent, sodium) for which a lower intake is desirable.

[0050] Here, if the user information data includes information about the actual state of the user's physical activity (specifically, exercise or daily activities), the evaluation unit 102c may use a machine learning model to obtain an activity evaluation score related to the quality of the user's current physical activity based on the user information data. The activity evaluation score is a score that determines the quality of the physical activity and may be calculated from multiple physical activity-related indicators according to the user's desired health outcomes (e.g., the physical activity the user wants to achieve). For example, the activity evaluation score may be the sum of a first evaluation score based on the ratio of actual exercise time to recommended exercise time, a second evaluation score based on exercise intensity, and a third evaluation score based on exercise frequency.

[0051] Furthermore, if the user information data includes information about the user's actual sleep habits, the evaluation unit 102c may use a machine learning model to acquire a sleep evaluation score related to the user's current sleep quality based on the user information data. The sleep evaluation score is a score for determining the quality of sleep and may be calculated from multiple sleep-related indices according to the user's desired health outcomes (e.g., sleep goals the user wants to achieve). For example, the sleep evaluation score may be the sum of a first evaluation score based on the ratio of actual sleep time to recommended sleep time, a second evaluation score based on the ratio of deep sleep, and a third evaluation score based on sleep interruptions.

[0052] For an example of predicting target data from text data using machine learning with the second acquisition unit 102b and the evaluation unit 102c, see Non-Patent Document 5. Non-Patent Document 5: Sagarika Shreevastava and Peter W. Foltz, “Detecting Cognitive Distortions from Patient-Therapist Interactions,” Proceedings of the Seventh Workshop on Computational Linguistics and Clinical Psychology, pages 151-158, June 11, 2021

[0053] The generation / calculation execution unit 102d generates intervention measure data (e.g., text data) related to an intervention measure that can be added to the acquired intervention measures based on the intervention measure data acquired by the second acquisition unit 102b and the diet evaluation score acquired by the evaluation unit 102c, and calculates the amount of change in the user's diet evaluation score due to the addable intervention measure. In this embodiment, the generation / calculation execution unit 102d has two functions, generation and calculation, but it is sufficient if it has at least the generation function. Specifically, it is sufficient if the generation / calculation execution unit 102d has at least the function of generating intervention measure data related to an intervention measure based on the diet evaluation score acquired by the evaluation unit 102c (the same applies to the activity evaluation score and the sleep evaluation score described above).

[0054] The generation / calculation execution unit 102d may prioritize interventions based on the actual dietary habits, health events, current dietary assessment score, the effects of interventions on the dietary assessment score, and the framework, and may propose "measures that will result in a significant change in the dietary assessment score when implemented" by considering conditions such as "proposing measures that will result in a large difference in the dietary assessment score" and "not proposing interventions that are similar to the currently adopted framework." A detailed description of the process related to this proposal will be provided later with reference to FIGS. 11 to 14. Note that, if a sufficient amount of data related to proposals and results has been accumulated, the processes of generating intervention measure data and calculating the amount of change may be performed using machine learning. Furthermore, this proposal process is not limited to cases involving dietary habits, but can also be applied to cases involving physical activity or sleep.

[0055] The generation / calculation execution unit 102d may calculate the change in the first evaluation score, the change in the second evaluation score, and the change in the third evaluation score, and calculate the sum of the calculated changes as the change in the diet evaluation score (the same applies to the activity evaluation score and sleep evaluation score mentioned above).

[0056] A specific description of the generation / calculation execution unit 102d will be given later with reference to FIG.

[0057] The output processing unit 102e outputs the intervention measure data generated by the generation / calculation executing unit 102d and the amount of change calculated by the generation / calculation executing unit 102d. A specific description of the output processing unit 102e will be given later with reference to FIG.

[0058] FIG. 4 is a diagram showing an example of content information data stored in the storage unit 106. As shown in FIG.

[0059] The content information data includes the following data: Content ID Text title (e.g., the title of a paper, book, or publicly known document) Body (e.g., the body of a paper, book, or publicly known document (e.g., abstract, method, or results))

[0060] The content information data may be linked to intervention measure category data, which includes the following data and which is acquired by the second acquisition unit 102b, as shown in Fig. 4: Intervention content (for example, a text summarizing the intervention content) Expected effect of the intervention content / Category of the continuity of the effect expected from the intervention content (for example, a category indicating the strength of continuity) Category of the range (effect range) of the effect expected from the intervention content (for example, "global", "local", etc.) Framework (category of measure) Effect amount of the effect expected from the intervention content / Effect of the intervention measure on the dietary assessment score

[0061] The second acquisition unit 102b may acquire the intervention content, the expected effect of the intervention content, the category of the sustainability of the expected effect of the intervention content, and the range and framework of the expected effect of the intervention content using a large-scale language model (LLM (Large Language Models)) such as a Generative Pre-trained Transformer (GPT). The control unit 102 may perform Retrieval-Augmented Generation (RAG) using, for example, "data stored in the memory unit 106," "data acquired by the acquisition unit (the first acquisition unit 102a and the second acquisition unit 102b)," and "data resulting from the user's implementation of the intervention measure proposed by the measure proposal device 100 (the intervention measure recommended to be implemented additionally)." This can improve the quality of data acquired from the user (the quality of the answers from the user).

[0062] The second acquisition unit 102b may acquire the effect size of the expected effect of the intervention content and the effect of the intervention measure on the dietary assessment score using a machine learning model.

[0063] FIG. 5 is a diagram showing an example of user information data stored in the storage unit 106. As shown in FIG.

[0064] The user information data includes the following data: User ID Survey region category (e.g., Asia, Europe, etc.) Survey country name and age category (e.g., teens, twenties, etc.) Gender category (e.g., male, female, etc.) Dietary habits (e.g., whether or not the user has a particular dietary habit) User's food literacy (e.g., knowledge about nutrition, cooking skills, general food-related skills, etc.)

[0065] The user information data may be linked to meal quality evaluation score data, including the user's current meal evaluation score, acquired by the evaluation unit 102c, as shown in FIG.

[0066] The first acquiring unit 102a may acquire the survey area category, the name of the survey country, the age category, the gender category, and the actual eating habits using a large-scale language model such as GPT.

[0067] The evaluation unit 102c may obtain the user's current meal evaluation score using a machine learning model.

[0068] The content information data (see FIG. 4) and user information data (see FIG. 5) stored in the storage unit 106 may be classified into the above frames and may be mainly used as learning data for the prediction model.

[0069] The memory unit 106 may store academic papers, research cases (e.g., nudge interventions regarding take-home sweeteners), publicly known literature, non-public information (e.g., research findings acquired and accumulated within the company), promotional cases, or personal health records as intervention content.

[0070] FIG. 6 is a diagram showing an example of a method for obtaining a diet evaluation score related to diet quality.

[0071] The evaluation unit 102c may convert the text data into a meal evaluation score, as shown in Fig. 6(B). The evaluation unit 102c may calculate a meal evaluation score based on the meal record data, in accordance with a scoring algorithm, and by referring to a score table, as shown in Fig. 6(A).

[0072] FIG. 7 is a diagram showing an example of a method for obtaining the intervention content, framework, range of effect expected from the intervention content, effect size of the effect expected from the intervention content, and effect of the intervention measure on the dietary assessment score.

[0073] The second acquisition unit 102b may acquire the intervention content, the framework, and the range of the effect expected from the intervention content by the procedure shown in Fig. 7. The second acquisition unit 102b may acquire the effect amount of the effect expected from the intervention content and the effect of the intervention measure on the dietary evaluation score by the procedure shown in Fig. 7.

[0074] FIG. 8 is a diagram showing an example of the content information data acquired by the first acquiring unit 102a and the intervention measure category data acquired by the second acquiring unit 102b.

[0075] The second acquisition unit 102b may acquire each of the categories and frameworks of the intervention content, the expected effect of the intervention content, and the sustainability of the expected effect of the intervention content using a large-scale language model such as GPT.

[0076] The second obtaining unit 102b may obtain the effect size of the expected effect of the intervention content and the effect of the intervention measure on the dietary assessment score using a machine learning model.

[0077] FIG. 9 is a diagram showing an example of the user information data acquired by the first acquisition unit 102a and the meal quality evaluation score data acquired by the evaluation unit 102c.

[0078] The first acquisition unit 102a may acquire each of the data listed below using a large-scale language model such as GPT. The first acquisition unit 102a may acquire the data listed below from the user's responses to a questionnaire about food. The first acquisition unit 102a may acquire the data listed below from food-related data recorded by application software such as a food recording app. - Category of the user's region of residence (e.g., Asia, Europe, etc.) - Name of the user's country of residence and age category of the user (e.g., teens, twenties, etc.) - Gender category of the user (e.g., male, female, etc.) - Actual eating habits of the user (e.g., people who follow a typical diet) - Health aspects desired by the user (e.g., text about the eating behavior achieved by the user (e.g., "I want to live a balanced diet"), etc.) - Food literacy of the user (e.g., knowledge about nutrition, cooking skills, general food-related skills, etc.)

[0079] The evaluation unit 102c may obtain the user's current meal evaluation score using a machine learning model.

[0080] The first acquisition unit 102a may acquire content information data corresponding to the format shown in Fig. 8 and user information data corresponding to the format shown in Fig. 9 by using a large-scale language model, user conversations, etc. The acquired data may be used as training data as appropriate to update the model.

[0081] Figure 10 shows an example of additional intervention measure data generated by the generation / calculation execution unit 102d and output by the output processing unit 102e, and an example of change data in the predicted evaluation score of diet quality calculated by the generation / calculation execution unit 102d and output by the output processing unit 102e.

[0082] The generation / calculation execution unit 102d generates intervention measure data regarding intervention measures that can be added to the acquired intervention measures based on the intervention measure data acquired by the second acquisition unit 102b and the dietary evaluation score acquired by the evaluation unit 102c, and calculates the amount of change in the user's dietary evaluation score due to the addable intervention measures.

[0083] The generation / calculation execution unit 102d may prioritize interventions based on the actual dietary habits, health events, current dietary assessment score, the effects of interventions on the dietary assessment score, and the framework, and may propose "measures that will significantly change the dietary assessment score when implemented" by considering conditions such as "proposing measures that will result in a large difference in the dietary assessment score" and "not proposing interventions that are similar to the currently adopted framework." A detailed description of the process related to this proposal will be provided later with reference to FIGS. 11 to 14. Note that if sufficient data on proposed measures to the user and data on the results of the user's implementation of the proposed measures have been accumulated to perform machine learning, the processes of generating intervention measure data and calculating the amount of change may be performed using machine learning. Furthermore, this proposal process is not limited to cases where dietary habits are the target, but can also be applied to cases where physical activity or sleep is the target.

[0084] The generation / calculation executing unit 102d may calculate the amount of change in the first evaluation score, the amount of change in the second evaluation score, and the amount of change in the third evaluation score, and calculate the sum of the calculated amounts of change as the amount of change in the diet evaluation score (the same applies to the activity evaluation score and the sleep evaluation score described above). The intervention measure data generated by the generation / calculation executing unit 102d may be text data. The generation / calculation executing unit 102d may store the generated data and the calculated data in a single table as shown in the upper part of Figure 10.

[0085] The output processing unit 102e outputs the intervention measure data generated by the generation / calculation executing unit 102d and the amount of change calculated by the generation / calculation executing unit 102d. Note that the output processing unit 102e may output the intervention measure data and the amount of change to an information communication terminal owned by the user in a format as shown in the lower part of Figure 10.

[0086] Fig. 11 is a flowchart showing an example of a measure proposal process executed by the generation / calculation execution unit 102d. Fig. 12 is a diagram showing a first embodiment of the measure proposal process. Fig. 13 is a diagram showing a second embodiment of the measure proposal process. Fig. 14 is a diagram showing an example of similarity evaluation between texts.

[0087] First, the generation / calculation execution unit 102d detects the nutrient or food with the lowest score among the nutrients or foods for which dietary evaluation scores have been calculated (Step 1). Specifically, the generation / calculation execution unit 102d selects the nutrient or food with the lowest score among the nutrients or foods for which scores have been calculated as the nutrient or food for which the score will be improved. For example, in both the first and second embodiments, the score for vegetables is the lowest, at 10, so vegetables are selected as the nutrient or food for which the score will be improved (see the upper part of Figure 12 and the upper part of Figure 13).

[0088] Next, the generation / calculation execution unit 102d extracts an intervention measure (specifically, an intervention measure category) that will most improve the score for the nutrient or food detected in step 1 (step 2). Specifically, the generation / calculation execution unit 102d selects the intervention measure that results in the greatest change in the dietary assessment score, calculated for each intervention measure, for the nutrient or food detected in step 1. For example, in both the first and second embodiments, the intervention measure category numbered 1 ("default") results in the greatest change of 10 in the score for vegetables (corresponding to the nutrient or food detected in step 1), so the intervention measure category numbered 1 ("default") is selected (see the bottom of Figure 12 and the bottom of Figure 13).

[0089] Next, the generation / calculation execution unit 102d determines a frame (category) corresponding to the current effort from the actual dietary habits, determines whether the intervention measure (specifically, the intervention measure category) extracted in step 2 overlaps with the determined frame (category) (step 3), and if there is an overlap (step 4: Yes), excludes the overlapping frame (category) and extracts the intervention measure (specifically, the intervention measure category) that will most improve the score for the nutrient or food detected in step 1 (step 5). For example, in the first embodiment, there is no frame (category) corresponding to the current effort, so it is determined in step 3 that there is no overlap (see the upper part of Figure 12). Also, for example, in the case of the second embodiment, the frame (category) corresponding to the current initiative is "default," and the category extracted in step 2 is also "default," so it is determined in step 3 that there is overlap.Then, in step 5, except for intervention category number 1 ("default"), intervention category number 2 ("commitment") produces the largest change in score for vegetables (corresponding to the nutrients or foods detected in step 1), at 2, so intervention category number 2 ("commitment") is selected (see Figure 13).

[0090] Next, the generation / calculation execution unit 102d determines whether there are multiple interventions extracted in step 2 or step 5 that most significantly improve the score for the nutrient or food detected in step 1 (step 6). If there are multiple interventions (step 7: Yes), the generation / calculation execution unit 102d estimates the similarity (specifically, cosine similarity, etc.) between the extracted (selected) intervention (or interventions with the same amount of change) and the health event desired by the user, and extracts interventions with the highest similarity (step 8). For a specific example of the similarity estimation performed in step 8, see FIG. 14. For example, in both the first and second embodiments, multiple interventions are not extracted, so the process does not proceed to step 8.

[0091] Next, the generation / calculation execution unit 102d proposes the intervention measures extracted in step 2, step 5, or step 8 (step 9).

[0092] This concludes the description of an example of the policy proposal process executed by the generation / calculation executing unit 102d.

[0093] Hereinafter, the operation of the policy proposal device 100 will be described again with reference to FIG.

[0094] 3, the second acquisition unit 102b and the evaluation unit 102c may convert the text data stored in the storage unit 106 into structured data (learning data) to create a prediction model. The second acquisition unit 102b and the evaluation unit 102c may (1) convert the text data into structured data via a large-scale language model as data preprocessing, and (2) design a learning model that predicts necessary information (intervention frame (category) and dietary evaluation score) from the structured data.

[0095] 3, the second acquisition unit 102b and the evaluation unit 102c may apply new data to the created model to obtain a prediction result. The second acquisition unit 102b and the evaluation unit 102c may (1) use a large-scale language model to acquire structured data to be used in the prediction model from a new user (multiple conversation patterns may exist), and (2) process the acquired structured data to perform prediction using the trained model (acquisition of information on the intervention frame (category) and dietary evaluation score).

[0096] With regard to the "operation (operation 3) of proposing additional measures and calculating predicted scores from intervention measure category data and diet quality evaluation score data" in FIG. 3, the generation / calculation execution unit 102d may propose additional measures and output estimated values ​​of their effects based on the acquired prediction results. The generation / calculation execution unit 102d may consider additional interventions by referencing the category information obtained in operation 2 and outputting frames (categories) that can be added. The generation / calculation execution unit 102d may quantify the effects of the additional factors in terms of diet evaluation scores by calculating the amount of change in the diet evaluation score due to the intervention.

[0097] 3, the output processing unit 102e may output information required for the created learning model. Specifically, the output processing unit 102e may output the intervention and frame (category) generated by the generation / calculation unit 102d, and the amount of change in dietary evaluation score expected from the intervention.

[0098] This embodiment has been described above, but by repeatedly outputting additional intervention measures and predicted score changes, and then inputting the consumer's eating behavior data, the following effects (1) to (3) can be obtained: (1) The accuracy of the output is improved (the prediction accuracy of the dietary quality evaluation score when the suggested eating behavior is implemented is improved). (2) The variety of the output is increased (among the eating behaviors included in the newly acquired text, newly suggested behaviors that are highly effective in improving dietary quality are presented). (3) It becomes possible to realize useful suggestions even for consumers to whom general eating behavior suggestions do not apply.

[0099] Furthermore, while this embodiment proposes personalized behavioral change measures to improve the quality of dietary life, for example, this embodiment may be applied to obtain "suggestions for product or service development" or "suggestions for promoting the use of developed products or services" within a company after collecting data from ordinary consumers, and services that can be realized by this embodiment may be provided to companies or governments. While this embodiment proposes personalized behavioral change measures to improve the quality of dietary life, personalized behavioral change measures to improve the quality of physical activity or sleep may also be proposed.

[0100] [3. Other Embodiments] Although the embodiments of the present invention have been described above, the present invention may be embodied in various different embodiments other than those described above within the scope of the technical concept set forth in the claims.

[0101] For example, among the processes described in the embodiments, all or part of the processes described as being performed automatically can be performed manually, or all or part of the processes described as being performed manually can be performed automatically using known methods.

[0102] In addition, the processing procedures, control procedures, specific names, registered data for each process, information including parameters such as search conditions, screen examples, and database configurations shown in the above documents and drawings can be changed as desired unless otherwise specified.

[0103] Furthermore, with regard to each device, the components shown in the drawings are functional concepts, and do not necessarily have to be physically configured as shown in the drawings.

[0104] For example, all or any part of the processing functions of the policy proposal device 100, particularly the processing functions performed by the control unit 102, may be implemented by a CPU or GPU and a program interpreted and executed by the CPU or GPU, or may be implemented as hardware using wired logic. The program is recorded on a non-transitory computer-readable recording medium containing programmed instructions for causing the information processing device to execute the processes described in the embodiments, and is mechanically read by the policy proposal device 100 as needed. That is, the storage unit 106, such as a ROM or HDD, stores a computer program for working with the OS to issue instructions to the CPU or GPU and perform various processes. The computer program is executed by being loaded into RAM and cooperates with the CPU or GPU to form the control unit.

[0105] In addition, this computer program may be stored in an application program server connected to the policy proposal device 100 via any network, and all or part of it may be downloaded as needed.

[0106] Furthermore, the policy proposal program for causing an information processing device to execute the policy proposal method described in this embodiment may be stored in a non-transitory computer-readable recording medium, or may be configured as a program product. Here, the term "recording medium" includes any "portable physical medium" such as a memory card, USB memory, SD card, flexible disk, magneto-optical disk, ROM, EPROM, EEPROM, CD-ROM, MO, DVD, and Blu-ray (registered trademark) Disc.

[0107] Furthermore, a "program" is a data processing method written in any language or description method, regardless of the format, such as source code or binary code. Note that a "program" is not necessarily limited to a single program, but also includes programs that are distributed as multiple modules or libraries, or programs that achieve their functions by working together with other programs, such as an OS. Note that the specific configurations and reading procedures for reading a recording medium in each device shown in the embodiments, as well as the installation procedures after reading, can use well-known configurations and procedures.

[0108] The various databases stored in the memory unit 106 are storage means such as memory devices such as RAM and ROM, fixed disk devices such as hard disks, flexible disks, and optical disks, and store various programs, tables, databases, and web page files used for various processes and providing websites.

[0109] The policy proposal device 100 may be configured as an information processing device such as a known personal computer or workstation, or may be configured as the information processing device to which any peripheral device is connected. The policy proposal device 100 may also be realized by installing software (including programs, data, etc.) that causes the information processing device to implement the policy proposal method described in this embodiment.

[0110] Furthermore, the specific form of distribution and integration of the devices is not limited to that shown in the drawings, and all or part of them can be functionally or physically distributed and integrated in any unit depending on various additions or functional loads. In other words, the above-mentioned embodiments can be implemented in any combination, or embodiments can be implemented selectively.

[0111] The present invention is particularly useful in services that provide behavioral change measures to improve the quality of life (especially dietary habits, physical activity, or sleep).

[0112] REFERENCE SIGNS LIST 100 Policy proposal device 102 Control unit 102a First acquisition unit 102b Second acquisition unit 102c Evaluation unit 102d Generation / calculation execution unit 102e Output processing unit 104 Communication interface unit 106 Storage unit 108 Input / output interface unit 112 Input device 114 Output device

Claims

1. A policy proposal device comprising: a first acquisition unit that acquires lifestyle data relating to a user's lifestyle; an evaluation unit that uses a machine learning model to acquire an evaluation score relating to the user's current quality of lifestyle based on the lifestyle data acquired by the first acquisition unit; a generation unit that generates intervention policy data relating to an intervention policy based on the evaluation score acquired by the evaluation unit; and an output processing unit that outputs the intervention policy data generated by the generation unit.

2. The policy proposal device according to claim 1, wherein the lifestyle data relates to the user's actual eating habits, and the evaluation score is a dietary evaluation score relating to the quality of meals.

3. The policy proposal device described in claim 2, characterized in that the dietary evaluation score is a score that determines the quality of a diet and is a score calculated from multiple types of nutrients or foods according to the health events desired by the user.

4. The policy proposal device described in claim 3, characterized in that the dietary evaluation score is the sum of a first evaluation score based on the content in the diet of a first nutrient that is preferable to be consumed in large amounts, a second evaluation score based on the content in the diet of a first food ingredient that is preferable to be consumed in large amounts, and a third evaluation score based on the content in the diet of a second nutrient that is preferable to be consumed in small amounts.

5. The measure proposal device according to claim 4, wherein the first nutrient is protein, the first food ingredient is a vegetable, and the second nutrient is saturated fatty acid and an amount of salt equivalent or sodium.

6. The measure suggestion device according to claim 5, wherein the health event is an eating behavior that the user wishes to achieve.

7. The policy proposal device according to claim 1, wherein the lifestyle data relates to the actual physical activity of the user, and the evaluation score is an activity evaluation score relating to the quality of the physical activity.

8. The policy proposal device described in claim 7, characterized in that the activity evaluation score is a score that determines whether physical activity is good or bad, and is a score calculated from multiple types of physical activity-related indicators according to the health events desired by the user.

9. The policy proposal device described in claim 8, characterized in that the activity evaluation score is the sum of a first evaluation score based on the ratio of actual exercise time to recommended exercise time, a second evaluation score based on exercise intensity, and a third evaluation score based on exercise frequency.

10. The measure suggestion device according to claim 9, wherein the health event is a physical activity that the user wants to achieve.

11. The policy proposal device according to claim 1, wherein the lifestyle data relates to the user's actual sleep habits, and the evaluation score is a sleep evaluation score relating to the quality of sleep.

12. The measure proposal device described in claim 11, characterized in that the sleep evaluation score is a score that determines the quality of sleep and is a score calculated from multiple types of sleep-related indicators according to the health events desired by the user.

13. The measure proposal device described in claim 12, characterized in that the sleep evaluation score is the sum of a first evaluation score based on the ratio of actual sleep time to recommended sleep time, a second evaluation score based on the ratio of deep sleep, and a third evaluation score based on sleep interruptions.

14. The measure suggestion device according to claim 13, wherein the health event is a sleep goal that the user wishes to achieve.

15. A measure proposal device as described in claim 6, 10 or 14, further comprising a second acquisition unit that acquires intervention measure data regarding the user's first intervention measure for the health event desired by the user, and the generation unit generates intervention measure data regarding a second intervention measure that can be added to the first intervention measure based on the intervention measure data acquired by the second acquisition unit and the evaluation score acquired by the evaluation unit.

16. The policy proposal device described in claim 15, further comprising a calculation unit that calculates the amount of change in the evaluation score of the user due to the second intervention policy based on the intervention policy data acquired by the second acquisition unit and the evaluation score acquired by the evaluation unit, and the output processing unit further outputs the amount of change calculated by the calculation unit.

17. The policy proposal device described in claim 16, characterized in that the calculation unit calculates the amount of change in the first evaluation score, the amount of change in the second evaluation score, and the amount of change in the third evaluation score, and calculates the sum of the calculated amounts of change as the amount of change in the evaluation score.

18. The policy proposal device described in claim 17, characterized in that the first acquisition unit acquires the actual living situation data using a large-scale language model, and the second acquisition unit acquires the intervention policy data using a large-scale language model.

19. The measure proposal device according to claim 18, wherein the intervention measure data generated by the generation unit is text data.

20. A method for proposing measures, comprising: a first acquisition step in which a first acquisition unit acquires lifestyle data relating to a user's lifestyle; an evaluation step in which an evaluation unit acquires an evaluation score relating to the user's current quality of lifestyle using a machine learning model based on the lifestyle data acquired in the first acquisition step; a generation step in which a generation unit generates intervention measure data relating to an intervention measure based on the evaluation score acquired in the evaluation step; and an output processing step in which an output processing unit outputs the intervention measure data generated in the generation step.

21. A policy proposal program that causes a computer to function as: a first acquisition means for acquiring lifestyle data relating to a user's lifestyle; an evaluation means for using a machine learning model to acquire an evaluation score relating to the user's current quality of lifestyle based on the lifestyle data acquired by the first acquisition means; a generation means for generating intervention policy data relating to an intervention policy based on the evaluation score acquired by the evaluation means; and an output processing means for outputting the intervention policy data generated by the generation means.

22. A method for creating a trained model by using learning data that associates lifestyle data related to a user's lifestyle with an evaluation score related to the quality of the user's current lifestyle, and executing a learning process for the parameters of a machine learning model that outputs an evaluation score when lifestyle data is input.

23. A trained model for outputting an evaluation score regarding the quality of a user's current living situation when actual living situation data regarding the user's living situation is input, wherein the parameters of the trained model are trained using training data that associates actual living situation data with evaluation scores, and the trained model is characterized by causing a computer to accept actual living situation data of a certain user as input, perform calculations based on the parameters on the input actual living situation data, and output an evaluation score.

24. A policy proposal device comprising: a first acquisition unit that acquires lifestyle data related to a user's lifestyle; an evaluation unit that uses a machine learning model to acquire an evaluation score related to the user's current quality of lifestyle based on the lifestyle data acquired by the first acquisition unit; a generation unit that generates intervention measure data related to an intervention measure based on the evaluation score acquired by the evaluation unit; and an output processing unit that outputs the intervention measure data generated by the generation unit, wherein the machine learning model is a trained model created by the creation method described in claim 22.

Images