Information processing program, information processing method, and information processing apparatus

Abstract

To improve the estimation accuracy of an action factor(s).SOLUTION: An information processing apparatus 100 generates action features showing the feature quantity related to the action(s) of each of plural persons based on the action plan(s) of each of the plural persons upon learning training (S1201). A machine learning model for estimating the action factor(s) of each object person is trained by using the feature quantity of the generated action features and correct answer information related to the action factor(s) of each of the plural persons. Plan data 111 on the action plan(s) related to the action(s) of the object person are inputted on the trained machine learning model upon estimating the action factor(s) of the object person to estimate the action factor(s) of the object person (S1202). Even when an action plan(s) caused by a factor(s) other than an action factor(s) as the inner state of each person is included in the action plan(s), reduction in the estimation accuracy of the action factor(s) is suppressed by the training using the generated action features. The estimated action factor(s) has a degree (high and low), and encourages a suitable action change on each object person depending on the degree of the estimated action factor(s) (S1203).SELECTED DRAWING: Figure 12

Description

[Technical Field]

[0001] The present invention relates to an information processing program, an information processing method, and an information processing device. [Background technology]

[0002] There is a technology that estimates the factors (behavioral factors) that determine the state of change over time in the behavior of a target person (behavioral change). Behavioral factors correspond to psychological (internal) characteristics such as the motivation and self-efficacy of the target person that make it easier for the target person to take the first step or feel like continuing that behavior. Estimating behavioral factors encourages the target person to make appropriate behavioral changes, and is applied to various support services such as health exercise. The target person's behavior can be detected as behavioral data, for example, from sensors on the target person's smartphone or the operation of a PC application. By using machine learning to learn the relationship between the behavioral data and behavioral factors of multiple people and creating a model, it is possible to use the created model to estimate the behavioral factors of the target person's new behavior.

[0003] Prior art includes, for example, a system that displays the amount of exercise and goal achievement status for a user-selected activity event based on information on the amount of exercise performed during the activity event, such as a predetermined sport, and supports rehabilitation planning. Another system creates an exercise plan that achieves a target heart rate and frequency without imposing a physical burden based on responses to a physical activity ability questionnaire, medication information, cardiopulmonary function classification results, exercise test results, etc. Another system uses the user's inclinations, preferences, and basic attributes such as age and gender as behavioral factors, and estimates factors that will induce a new user's purchasing behavior by learning the correlation between a questionnaire on the preferences of unspecified users and a series of behavioral data such as purchases, website browsing, and movement data (see, for example, Patent Documents 1 to 3 listed below). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2020-149321 [Patent Document 2] Japanese Patent Publication No. 2023-159553 [Patent Document 3] Japanese Patent Application Publication No. 2023-97496 Summary of the Invention [Problem to be solved by the invention]

[0005] However, if the behavioral data used for training includes data on behavior caused by factors other than behavioral factors, the estimation accuracy of the machine learning model may decrease.

[0006] In one aspect, the present invention aims to improve the accuracy of estimating behavioral factors. [Means for solving the problem]

[0007] According to one embodiment, an information processing program, an information processing method, and an information processing device are proposed, which acquire a behavioral plan indicating the planned behavior of each of a plurality of individuals, generate behavioral features indicating characteristic quantities related to the behavior of each of the plurality of individuals based on the acquired behavioral plan, and train a machine learning model that estimates the behavioral factors of a subject using the characteristic quantities of the generated behavioral features and correct answer information related to the behavioral factors of each of the plurality of individuals. [Effects of the Invention]

[0008] According to one aspect, it is possible to improve the accuracy of estimating behavioral factors. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is an explanatory diagram illustrating an example of information processing according to an embodiment. [Figure 2] FIG. 2 is an explanatory diagram of behavioral factor estimation according to the prior art. [Figure 3] FIG. 3 is an explanatory diagram for estimating behavioral factors from the behavioral data of a subject based on exercise log data. [Figure 4] FIG. 4 is a chart showing examples of behavioral factors estimated in the embodiment. [Figure 5] Figure 5 is a chart showing example questions from a psychological questionnaire that serve as correct answer data when learning a model that estimates behavioral factors. [Figure 6] FIG. 6 is a chart showing an example of information in an exercise task list. [Figure 7] FIG. 7 is a chart showing an example of information in the behavioral feature group list. [Figure 8] FIG. 8 is an explanatory diagram illustrating the quantification of behavioral features based on log data of exercise tasks. [Figure 9] FIG. 9 is a block diagram illustrating an example of the hardware configuration of an information processing device. [Figure 10] FIG. 10 is a flowchart showing an example of processing performed by the information processing device during model learning. [Figure 11] FIG. 11 is a flowchart showing an example of processing performed by the information processing device when estimating behavioral factors. [Figure 12] FIG. 12 is an explanatory diagram of the estimation of behavioral factors according to the present invention and the actions taken in response to the estimated behavioral factors of the subject. [Figure 13] FIG. 13 is an explanatory diagram of actions corresponding to estimated behavioral factors of a subject according to the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0010] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an information processing program, an information processing method, and an information processing device according to embodiments of the present invention will be described in detail with reference to the accompanying drawings.

[0011] (An example of an information processing method according to an embodiment) 1 is an explanatory diagram showing an example of information processing according to an embodiment. The information processing method according to the embodiment estimates behavioral factors of a target person's behavior using, for example, a machine learning model. In the following description, a person for whom a new behavioral factor is to be estimated is referred to as a "target person."

[0012] The behavioral factors are factors that have a large influence on behavioral change and correspond to the psychological (internal) state of a person that is a factor related to the person's behavior. The subject whose behavioral factors are to be estimated is, for example, a person who is undergoing a predetermined exercise training.

[0013] The information processing method according to the embodiment performs the following steps: 1. training a machine learning model; and 2. estimating behavioral factors using the created machine learning model. 1. Training a machine learning model involves the following steps: 1-1. Generate internal features (behavioral features) regarding the behavior of each of a plurality of people based on a plan (behavioral plan) regarding the behavior of each of a plurality of people. For example, when generating behavioral features, labeling is performed on the behavior plan. For example, labeling involves associating a load (Level) with each task of the planned exercise. Furthermore, in the process of generating behavioral features, feature amounts (behavioral features) may be generated based on the plan and the subject's implementation results for the plan. In this case, the implementation rate (achievement level) of the plan can be calculated as described below. The feature amount is a value having a vector based on the numerical values ​​included in the plan and the numerical values ​​included in the implementation results for the plan.

[0014] 1-2. Using the generated features (behavioral features) and correct information (psychological questionnaire) regarding the factors behind the behavior of each of multiple individuals, a machine learning model is trained (learning process) to estimate the factors behind the behavior of the target person. In addition, in the process of correcting and training features based on the difference between the plan and the implementation results, the machine learning model may be trained using the corrected features and information.

[0015] 2. When estimating behavioral factors using the created machine learning model, the following process is carried out. The target person's behavior plan is input into a machine learning model trained using internal characteristics (behavioral characteristics) related to the behavior of each of multiple individuals, which are generated based on the plans for each of the multiple individuals' behavior, and correct answer information related to the factors behind the behavior of each of the multiple individuals, and the factors behind the target person's behavior are estimated.

[0016] The information processing method according to the embodiment can be realized by the information processing device 100, such as a server, executing a program, or by accessing the cloud.

[0017] First, a description will be given of a schematic configuration of an information processing device 100 shown in Fig. 1. In the following embodiment, an example of application to a service that supports subjects who subscribe to a health exercise service in continuing healthy behavior will be described.

[0018] The information processing device 100, using the functions of Fig. 1 described below, supports a subject who has been exercising for a certain period of time by taking actions appropriate to the behavioral factors estimated after the certain period has elapsed. The information processing device 100 quantifies, as the subject's behavioral characteristics, trends and deviations in the load, frequency, number of times, duration per session, etc., related to the exercise actually performed by the subject in relation to the exercise action plan. The information processing device 100 then uses the behavioral characteristics as input data for a machine learning model (behavioral factor estimation model).

[0019] The information processing device 100 then trains (learns) a model (behavioral factor estimation model) that estimates the behavioral factors of the subject, using the input behavioral features and the aggregated results of a psychological questionnaire 125 of information on the behavioral factors of each of the multiple persons. Furthermore, using the trained behavioral factor estimation model, the information processing device 100 estimates and outputs multiple types of behavioral factors of the subject, based on the correlation between the numerical values ​​of the behavioral features and the values ​​of the psychological questionnaire obtained separately.

[0020] The information processing device 100 registers a group of motor tasks and the degree of load (difficulty of performing) of each task in advance as an motor task list. The information processing device 100 may also represent a plan for motor tasks to be performed over a certain period of time in the future as a vector, calculate the bias and center of gravity indicated by the vector, and quantify the behavioral feature using the calculated bias and center of gravity. Details will be described later, but for example, the implementation rate (achievement level) is calculated by dividing the results of the implementation of the plan for a certain period of time. The center of gravity is also calculated from the calculated implementation rate vector. By calculating the center of gravity, the subject's level of challenge in a behavioral feature, for example, the "challenge level" behavioral feature, can be quantified.

[0021] The following describes each function of the information processing device 100 shown in Fig. 1. The information processing device 100 includes the functions of a log data preprocessing unit 101, a vector conversion unit 102, a behavioral feature calculation unit 103, a weighting correction unit 104, a machine learning unit 105, and a behavioral factor estimation model (machine learning model) 106.

[0022] 1 acquires log data 110 related to exercise tasks and an exercise task list 121, and performs preprocessing to calculate behavioral features. The log data 110, which will be described in detail later, includes plan data 111 for a certain period of time in the past (e.g., one month). It may also include plan implementation result data 112.

[0023] The plan data 111 is data on planned exercise. The plan data 111 includes, for example, information on the number of tasks (0: no plan / 1 to n: plan, n is the number of planned tasks) indicating which exercise is planned for each planned day at each level (Level 1 to Level 5) of load (difficulty) of a plurality of exercises.

[0024] The implementation result data 112 is data indicating the results of the subject actually implementing the exercise plan (task) shown in the plan data 111. For example, the implementation result data 112 is data indicating whether or not the subject implemented each task planned as a plurality of plan data 111 (0: not implemented / 1 to n: implemented, n is the number of tasks implemented). The implementation result data 112 is created, for example, by the subject himself / herself by inputting data at the end of each planned day (such as in the evening).

[0025] As a preprocessing step for calculating behavioral features, the log data preprocessing unit 101 acquires an exercise task list 121 and sorts the log data 110 for a predetermined period (e.g., the past seven days) for calculating the behavioral features. The exercise task list 121 lists levels (Level 1 to Level 5) by load and the exercise content (tasks) for each level (see FIG. 6).

[0026] The log data preprocessing unit 101 refers to the exercise task list 121 and performs preprocessing such as labeling of the plan data 111 of the acquired log data 110 and the plan implementation result data 112. For example, the log data preprocessing unit 101 sets each day of a predetermined period (the past seven days) vertically and the level of each load horizontally, and generates labeled data arranged in a table format where the intersections of the vertical and horizontal axes indicate each task (see FIGS. 8(a) and (b)).

[0027] The vector conversion unit 102 performs vector conversion for each exercise load on the data preprocessed by the log data preprocessing unit 101, i.e., the plan data 111 for a predetermined period (past seven days) and the implementation result data 112. The distances between tasks in the tabular plan data 111 after sorting correspond to vectors.

[0028] The behavioral feature calculation unit 103 refers to the behavioral feature group list 123, calculates the center of gravity of the vector, and calculates various behavioral features as numerical values. Specific examples of calculating behavioral features will be described later. The behavioral feature group list 123 is a list of examples of behavioral features prepared in advance (see FIG. 7). Representative behavioral features include, for example, achievement level "achieving a plan" and challenge level "taking on a task with high load or difficulty."

[0029] The behavioral features calculated by the behavioral feature calculation unit 103 are used, for example, to determine what kind of exercise a health exercise promotion manager or the like will encourage a subject to do on the next day of a predetermined period (the past seven days). The behavioral features calculated by the behavioral feature calculation unit 103 include, for example, the level of challenge for exercise indicated by the exercise load tendency and the level of achievement indicated by the exercise plan achievement tendency.

[0030] The weighting correction unit 104 shown in FIG. 1 applies a predetermined weight to the calculated behavioral features, thereby improving the accuracy of machine learning by the machine learning unit 105. The weighting correction unit 104 weights the behavioral features based on the difference between the numerical value of the behavioral feature calculated by the behavioral feature calculation unit 103 and the correction data 124. The correction data 124 is, for example, information such as the subject's past exercise results (e.g., results from one year ago) and the current level of exercise habits. The weighting correction by the weighting correction unit 104 using the correction data 124 may be performed, for example, once a year, rather than every time a correction is performed.

[0031] The machine learning unit 105 trains a machine learning model (behavioral factor estimation model 106) that estimates the factors behind the behavior of a subject using the input behavioral characteristics and the aggregated results of a psychological questionnaire 125 containing information about the factors behind the behavior of each of multiple subjects.

[0032] The machine learning model (behavioral factor estimation model 106) trained by the machine learning unit 105 estimates and outputs multiple types of behavioral factors of the subject and the degree value of each behavioral factor (for example, two values ​​of high / low) based on the correlation between the numerical values ​​of the behavioral features and the values ​​of the separately acquired psychological questionnaire 125. As will be described in detail later, the behavioral factor estimation model 106 learns the relationship between the behavioral features and the behavioral factors, and thereby outputs, for example, the degree of self-efficacy (degree of confidence) as either high or low.

[0033] (Regarding the prior art) Figure 2 is an explanatory diagram of behavioral factor estimation using conventional technology. Behavioral factors are estimated by learning the relationship between the subject's past exercise logs (how many times and how frequently they performed various types of exercise) and the "behavioral factors" quantified in a psychological questionnaire 125.

[0034] 2(a) shows an example of processing during learning. A behavior change application 201 installed on a smartphone held by a subject accumulates and stores a log 202 of exercises performed by the subject. The log 202 includes behavioral data on the types and number of exercises performed by the subject (step S201). The subjects' behavioral factors are divided into those with high behavioral factors and those with low behavioral factors (step S202).

[0035] During learning, the behavioral data of multiple people is compared with the correct values ​​of the behavioral factors (psychological questionnaire 125), and the correspondence between the behavioral data and the behavioral factors is trained to generate a model 203 (step S203). In the example of FIG. 2(a), the horizontal axis represents certain behavioral data 1, and the vertical axis represents other behavioral data 2, and the correct values ​​of the behavioral factors of multiple people are obtained based on a psychological scale such as a five-point scale. The correct values ​​are set to two values, high or low, divided by a predetermined average standard.

[0036] In conventional technology, for example, to estimate binary values ​​(high / low) of behavioral factors from behavioral data, logistic regression is performed using the behavioral data as the explanatory variable and the correct value as the objective variable. For example, two pieces of behavioral data 1 and 2 are selected as the horizontal and vertical axes and used as explanatory variables. Boundary formula D is calculated using logistic regression. Then, when a new subject to be estimated is plotted on a chart, the behavioral factor is estimated based on which area the subject is located in, with boundary formula D at the center. For example, the behavioral factor of a subject plotted in area R1 on one side of boundary formula D can be estimated to be "low," and the behavioral factor of a subject plotted in area R2 on the other side of boundary formula D can be estimated to be "high."

[0037] 2(b) shows an example of estimating behavioral factors using a model. Behavioral data of a subject is input to the trained model 203 (step S204). As a result, the model 203 outputs an estimation result of the subject's behavioral factors (for example, in the illustrated example, the behavioral factor is "low").

[0038] 3 is an explanatory diagram illustrating how behavioral factors are estimated from the behavioral data of a subject based on log data. For example, a device such as a smartphone operated by multiple people can acquire log data 110 of the people. The log data 110 includes information such as the type of exercise intensity, the number of times the exercise was performed, when the exercise was performed, and the elapsed time since the exercise was performed.

[0039] The information processing device receives the acquired log data 110 as input, extracts behavioral data a for each behavioral feature set in advance in a behavioral data extraction layer 302, and extracts an optimal combination pattern of the behavioral data a by machine learning 304, thereby estimating behavioral factors C. During learning, correct examples (correct answer data) of behavioral factors based on the questionnaire results Q of the psychological questionnaire 125 conducted in advance are referenced from a questionnaire result storage unit 306 that stores the questionnaire results Q of the psychological questionnaire 125.

[0040] In this way, by learning the relationship between the past behavioral data a of multiple people and the behavioral factors C, it becomes possible to estimate the behavioral factors C of the behavioral data of a new (unknown) subject using an information processing device. For example, the behavioral factors C to be estimated include the subject's exercise motivation, cooperation motivation, self-efficacy, etc.

[0041] However, in the construction of a learning model based on exercise logs that was previously implemented as described in the prior art, a. When behavioral factors caused the inability to perform the task (e.g., lack of confidence), b. It is not possible to distinguish between reasons unrelated to the behavioral factors (e.g., bad weather, staying at home working from home) and the fact that the exercise was not performed, and it is treated as the same as "not doing that exercise." This leads to incorrect learning (treating them as equivalent) when building a learning model, and the accuracy of the final estimation of behavioral factors also deteriorates.

[0042] When learning from a large amount of data on a single individual over a long period of time (for example, several years' worth), it may be possible to absorb factors other than behavioral factors and sudden events, but when trying to create a learning model from one month's worth of practical data on hundreds of individuals, the accuracy of estimating the subject's behavioral factors will deteriorate due to the extremely large noise in the model.

[0043] In the embodiment, to solve the above problem, machine learning is not simply performed based on log data of past exercise. In the embodiment, based on a future health behavior plan to be implemented by the subject for a predetermined future period (e.g., 7 days), the plan is learned by "quantifying trends and biases in load, frequency, and number of times as the subject's behavioral characteristics by labeling them," with the degree of exercise habit to date regarded as a baseline. According to the embodiment, machine learning can be performed on future plans without being affected by factors unrelated to behavioral factors, such as weather or teleworking, and by removing noise from unexpected events, thereby improving the accuracy of estimating behavioral factors.

[0044] (Examples of behavioral factors) Figure 4 is a chart showing examples of behavioral factors estimated in the embodiment. In the example shown in Figure 4, behavioral factors are classified into the following nine types, 1. to 9., based on theories of behavioral change defined in the following Technical Document 1 and on-site interviews with people who engage in exercise.

[0045] Technical Reference 1: Shoji Hayakawa and 7 others, Study on Factors Determining Elderly Care Prevention Behavior, Proceedings of the Multimedia, Distributed, Collaborative and Mobile Symposium 2022, pp.590-594

[0046] In the example shown in Figure 4, behavioral factors C are broadly categorized into three: cognitive factors, ability factors, and environmental factors. Cognitive factors include five types of behavioral factors C: (1) "knowledge" about the behavior, (2) "motivation" to perform the behavior, (3) "self-efficacy" about the behavior, (4) "prediction of the outcome" about the behavior, and (5) "time and cost" for the behavior. The content of each type, for example, (1) "knowledge" about the behavior refers to whether or not one has knowledge about the "necessary knowledge" to perform the behavior and the "impact of the behavior" (however, it is difficult to bring about behavioral change with knowledge alone).

[0047] In addition, there are two types of ability factors C: (6) "physical and intellectual ability" for the behavior and (7) "ease of incorporating the behavior into daily routine." For example, (6) "physical and intellectual ability" for the behavior refers to whether or not a person has sufficient "physical and intellectual ability" to perform the behavior.

[0048] In addition, there are two types of environmental factors: (8) "Observation of a role model" for that behavior, and (9) "Social support" for that behavior. For example, (8) "Observation of a role model" for that behavior means "there is someone close to you who is performing or succeeding in that behavior (such as an elderly person close to you who is in a similar situation)" and "you know about it (observational learning)."

[0049] (Example of a psychological questionnaire) Figure 5 is a diagram showing example questions from a psychological questionnaire that serve as correct answer data when training a model to estimate behavioral factors. The example questions q shown in Figure 5 use the SMSGSE disclosed in the following technical document 2 and the psychological questionnaire based on the Multi-Aspect Work Motivation Scale disclosed in the following technical document 3.

[0050] Technical Reference 2: Akiko Miyoshi, Development of the Scale of Personality-Specific Self-Efficacy as a Subjective Feeling (SMSGSE), Research in Developmental Psychology, vol.14(2), pp.172-179, 2003 Technical Reference 3: Hiroshi Ikeda, Yuta Morinaga, Development of a Multidimensional Work Motivation Scale in Japan, Industrial and Organizational Psychology Research, vol.30(2), pp.171-189, 2017

[0051] In the example question q shown in Figure 5, for example, three categories of behavioral factors are set: self-efficacy, cooperative motivation, and learning motivation, and multiple questions are asked of the subject based on each behavioral factor. For example, the multiple questions about the behavioral factor "self-efficacy" are "I feel like I can do most things without much effort," "No matter what situation I face, I feel like I can handle it well," ... "Even in the most difficult situations, I feel like I can get out of it." For example, correct answer data is generated based on the person's responses to each of these questions on a five-point scale (based on the level of achievement).

[0052] (Examples of processing by each function of an information processing device) Next, a specific example of processing performed by the information processing device 100 will be described. Details of the processing performed by the information processing device 100 shown in Fig. 1 will be described. As a preliminary preparation for the function of learning the above-mentioned "by quantifying trends and biases in load, frequency, and number of times as a person's behavioral characteristics," several dozen types of "exercise tasks" that can be performed on a daily basis are registered as a list.

[0053] 6 is a chart showing an example of information in an exercise task list. In the exercise task list 121, a plurality of exercise tasks (exercise content) are registered in association with a value (Level 1 to Level 5) for the degree of exercise load (how difficult it is). Level 1 (sitting on a chair and lying sideways for 5 seconds) is the least exercise load and can be easily performed, while Level 5 (20 minutes of jogging around the house) is the most exercise load and is difficult to perform.

[0054] The subject selects, for example, up to three tasks per day from the exercise task list 121 and creates a plan (plan data 111) of what exercise he or she wants to do for the next week (7 days).

[0055] The log data preprocessing unit 101 of the information processing device 100 acquires a one-week exercise task plan (plan data 111). The vector conversion unit 102 of the information processing device 100 expresses as a vector the number of planned exercises to be performed for each level of exercise load (Level), and the behavioral feature calculation unit 103 calculates the center of gravity of the vector.

[0056] Fig. 7 is a chart showing an example of information in a behavioral feature group list. The behavioral feature group list 123 shown in Fig. 7 shows examples of setting two representative behavioral features 123a: achieving (a plan) and taking on (a high-stress or high-difficulty task). In this embodiment, the subject's tendency is calculated for each of the degree of achievement of the exercise plan and the degree of challenge for the plan.

[0057] Other behavioral characteristics that can be set in the behavioral characteristics group list 123 include unevenness in the weeks when an action is taken and the weeks when an action is not taken, a plan in which the frequency of action and high load are concentrated in the early stages of the plan, a plan in which the frequency of action and high load are concentrated in the later stages of the plan, a constant frequency of action during the period (the load and frequency selected are constant during the period), etc.

[0058] The behavioral feature calculation unit 103 quantifies the behavioral features of the subject based on the behavioral features in the behavioral feature group list 123 and the output of the vector conversion unit 102. For example, if the center of gravity is close to Level 1 in the exercise task list 121, the behavioral feature calculation unit 103 determines that the subject has a "low level of the behavioral feature of taking on challenges." If the focus is on Level 5 of the Exercise Task List 121, it is judged that the "behavioral characteristic of challenging oneself is high."

[0059] Furthermore, the behavioral characteristic calculation unit 103 calculates the ratio of the implementation result of actual exercise to the exercise plan. For example, the behavioral characteristic calculation unit 103 calculates the implementation rate indicating the extent to which the exercise was implemented by dividing the implementation result of actual exercise by the exercise plan for one week (7 days). The behavioral characteristic calculation unit 103 also calculates the center of gravity of the calculated implementation rate vector.

[0060] As a result, the behavioral feature calculation unit 103 If the implementation rate is below a certain level, it is judged that the behavioral characteristics of achieving (the plan) are low. If the implementation rate is above a certain level and the focus is on Level 1, it is judged that "the behavioral trait of achievement is high, but the behavioral trait of challenge is low." If the implementation rate is above a certain level and the focus is on Level 5, it is judged that "the behavioral characteristics of achievement are high and the behavioral characteristics of challenge are also high."

[0061] Figure 8 is an explanatory diagram illustrating the quantification of behavioral characteristics based on exercise task log data. Figure 8(a) is a diagram of exercise plan data 111 planned by the subject himself / herself. The vertical axis of the plan data 111 represents each day of a week (7 days), and the horizontal axis represents the planned load (level). Tasks for each exercise load (level) are set in 35 squares (7 days x 5 (levels)). The bottom row shows the total number of tasks for each load (level) planned for the week. For example, the subject has planned two tasks with a load of level 1 and one task with a load of level 4 for Monday, August 1st. He also has planned one task with a load of level 1 and one task with a load of level 2 for Monday, August 2nd. In this table, vectors correspond to arrows between the planned tasks (values ​​other than 0).

[0062] FIG. 8(b) is a diagram of the implementation result data 112, showing the exercise tasks by load (Level) that the subject actually performed on each evening of one week (7 days). As with the planned data 111, the implementation result data 112 also has a total of 35 task counts set. The bottom row shows the total number of tasks by load (Level) for one week of actual exercise. Comparing the planned data 111 and the implementation result data 112, the difference is that the planned exercise tasks with a load of Level 1 were not performed on August 3, August 4, and August 6 (number of tasks: 0).

[0063] The log data preprocessing unit 101 preprocesses plan data 111 of exercise task log data 110 acquired for a predetermined period, and generates the tabulated plan data 111 shown in Fig. 8(a). Similarly, the log data preprocessing unit 101 preprocesses plan implementation result data 112 acquired for a predetermined period, and generates the tabulated implementation result data 112 shown in Fig. 8(b).

[0064] The behavioral feature calculation unit 103 converts the trend for the past seven days into one sample of data based on the plan data 111 in Fig. 8(a) and the implementation result data 112 in Fig. 8(b). For example, it performs a numerical calculation on the behavioral features to determine what kind of encouragement should be given to the subject eight days after the planned one week has elapsed (Monday, August 8th).

[0065] Specifically, the behavioral characteristic calculation unit 103 counts the trends of plans and implementation results for each exercise load (Level) over the past seven days. First, the behavioral characteristic calculation unit 103 calculates the challenge level (load trend) P using the following formula 1 based on the total number of plans for each load. P=(p1+p2+p3+p4+p5)=8+4+0+3+0=15 …(1) If you do three tasks every day, the maximum total value is 3 x 7 = 21.

[0066] Furthermore, the behavioral characteristic calculation unit 103 calculates the achievement level (implementation rate, plan achievement tendency) A using the following formula 2. The achievement level (implementation rate) A has a vector A, which has a predetermined bias and center of gravity, and the behavioral characteristic calculation unit 103 calculates the center of gravity from the implementation rate vector. A = (Result B / Plan P) = (b1 / p1, b2 / p2, …, b5 / p5) =(5 / 8,4 / 4,…,0 / 0) …(2)

[0067] When the function of the weighting correction unit 104 is used, the following weighting correction is performed on the numerical values ​​of the behavioral features calculated by the behavioral feature calculation unit 103. The weighting correction unit 104 holds, as correction data 124, data on the subject's past exercise results and the current level of exercise habit. For example, as the correction data 124, the weighting correction unit 104 calculates the subject's exercise overconfidence from the difference between actual results and planned results. In addition, the weighting correction unit 104 calculates the behavioral factors based on the implementation results of the plan. The weighting correction unit 104 performs periodic correction, for example, once a year, and corrects the numerical values ​​of the behavioral factors using the calculated overconfidence. The weighting correction is performed, for example, once a year, to improve the accuracy of machine learning, and correction reflecting actual results is not performed every time (for example, every week). The overconfidence does not have to be calculated using actual results from a year ago, but may also be obtained separately using a psychological questionnaire or the like.

[0068] As a correction method different from the above, the weighting correction unit 104 may multiply or add a coefficient that corrects the achievement level to a higher value when the calculation result of the achievement level A is 1, as the value of B (= P) becomes larger. In this case, the achievement level is corrected by regarding a case where a subject declares four tasks and completes four of them as having "achieved a more difficult task" than a case where a subject declares one task and completes one of them per day.

[0069] (Example of hardware configuration of information processing device) Fig. 9 is a block diagram showing an example of the hardware configuration of an information processing device. In Fig. 9, a general-purpose server can be used as the information processing device 100. The information processing device 100 includes a CPU (Central Processing Unit) 901 such as a processor, a memory 902, a network IF 903, a recording medium IF 904, and a recording medium 905. The components are connected to each other via a bus 900.

[0070] Here, the CPU 901 is a control unit that controls the entire information processing device 100. The CPU 901 may have multiple cores. The memory 902 includes, for example, a read-only memory (ROM), a random access memory (RAM), and a flash ROM. Specifically, for example, the flash ROM stores an OS program, the ROM stores application programs, and the RAM is used as a work area for the CPU 901. The programs stored in the memory 902 are loaded into the CPU 901, causing the CPU 901 to execute the coded processes.

[0071] The network IF 903 is connected to the network NW via a communication line and can be connected to an external computer via the network NW. The network IF 903 serves as an interface between the network NW and the inside of the device, and controls the input and output of data from the external computer. For example, a modem or a LAN adapter can be used as the network IF 903.

[0072] The recording medium IF 904 controls reading / writing of data from / to the recording medium 905 under the control of the CPU 901. The recording medium 905 stores the written data. Examples of the recording medium 905 include a magnetic disk such as an HDD, an optical disk, an SSD (Solid State Drive), and a USB (Universal Serial Bus) memory.

[0073] In addition to the above-described components, the information processing device 100 may also include, for example, an input device, a display, and the like.

[0074] Each function of the log data preprocessing unit 101 to the behavioral factor estimation model 106 shown in Fig. 1 can be realized by executing a program on the CPU 901 shown in Fig. 9. Each piece of data in the data storage unit (exercise task list 121, behavioral feature group list 123, correction data 124, and psychological questionnaire 125) shown in Fig. 1 can be stored using the memory 902 and recording medium 904 shown in Fig. 9. The function of acquiring the log data 110 by the log data preprocessing unit 101 shown in Fig. 1 and the function of outputting the behavioral factors estimated by the behavioral factor estimation model 106 can be realized using, for example, the network IF 903 shown in Fig. 9.

[0075] 1 can also use functions on the cloud via a network NW connection. In this case, the CPU 901 shown in Fig. 9 transmits data of calculated behavioral features to the cloud and receives data of behavioral factors from the cloud.

[0076] (Example of processing during model training) 10 is a flowchart showing an example of processing performed by the information processing device when the information processing device 100 learns (trains) the behavior factor estimation model 106. The following processing is executed by the CPU 901 (a control unit having the functions shown in FIG. 1) of the information processing device 100.

[0077] First, the information processing device 100 acquires planned data 111 of the exercise task of the subject as input to the device (step S1001). Next, the information processing device 100 extracts data for a predetermined period (one week) required for processing from the planned data 111 (step S1002).

[0078] Next, the information processing device 100 acquires the load level (Lv1 to Lv5) of each task from the exercise task list 121 (step S1003), and labels the load level to the extracted plan data for one week (7 days) (step S1004). By this labeling, an exercise load (Lv) is set for each task of the planned exercise in the plan data 111.

[0079] Next, the information processing device 100 converts the labeled plan data 111 into a vector for one week (step S1005). For example, the load (Lv) of each task in the one-week plan is summed.

[0080] Next, the information processing device 100 calculates the center of gravity of the data after vector conversion (step S1006). In the example shown in Fig. 10, the center of gravity of the vector is shifted to the side with a light load (Lv1).

[0081] Next, the information processing device 100 calculates the subject's exercise challenge level (step S1007). For example, if the calculated center of gravity is located on the side with the highest load (Level 5), the challenge level is determined to be the maximum "1," and if the center of gravity is located on the side with the lowest load (Level 1), the challenge level is determined to be the minimum "0." Here, the challenge level is determined as a binary value "1 / 0" corresponding to the binary value of the behavioral factor, high / low.

[0082] Then, the information processing device 100 inputs the value (numerical value) of the behavioral feature "level of challenge" into the model (behavioral factor estimation model 106) as output (step S1008), and ends the above processing.

[0083] (Example of processing when estimating behavioral factors) Fig. 11 is a flowchart showing an example of processing performed by an information processing device when estimating behavioral factors. An example of processing performed by the information processing device 100 when estimating the behavioral factors of a subject using the behavioral factor estimation model 106 learned through the processing in Fig. 10 will be described. The processing in Fig. 11 is executed by the CPU 901 (a control unit consisting of the functions shown in Fig. 1) of the information processing device 100, and includes processing for encouraging the subject to appropriate exercise based on the estimated behavioral factors.

[0084] First, the information processing device 100 acquires plan data 111 of a new subject's exercise task as input to the device (step S1101). Next, the information processing device 100 extracts one week's worth of plan data required for processing from the plan data 111, and calculates the vector center of gravity of the extracted one week's worth of plan data (step S1102). The processing in step S1102 is the same processing as during learning (steps S1002 to S1006 in FIG. 10). Through these vector conversions and calculations of the center of gravity, meanings are assigned to the plan data 111 at the time of planning, and the meanings are used as input for learning the model.

[0085] Next, the information processing device 100 inputs the behavioral features for which the vector centroids have been calculated into the model (behavioral factor estimation model 106) (step S1103). As a result, the information processing device 100 calculates the degree of behavioral factor estimation by the model (behavioral factor estimation model 106) as output (step S1104). For example, the information processing device 100 calculates and outputs an estimated score = x for the behavioral factor "challenge motivation."

[0086] The subsequent processing may be performed by the information processing device 100, or may be performed by an administrator or the like based on the output. For example, the information processing device 100 determines whether the estimated score x of challenge motivation is equal to or greater than a predetermined threshold (step S1105).

[0087] If the estimated score x of challenge motivation is equal to or greater than a predetermined threshold (step S1105: Yes), the information processing device 100 sends the subject a message encouraging him / her to exercise, for example, a message saying "Why not try a hard challenge?", two weeks after starting the exercise (step S1106). On the other hand, if the estimated score x of challenge motivation is less than the predetermined threshold (step S1105: No), the information processing device 100 does not send a message to the subject, or sends a standard message saying "Let's continue (your current exercise)" (step S1107). After processing step S1106 or step S1107, the information processing device 100 ends the above processing.

[0088] (Summary of the information processing method of the present invention) 12 is an explanatory diagram of the estimation of behavioral factors according to the present invention and the actions taken in response to the estimated behavioral factors of the subject. In the example of FIG. 12, the subject creates a lifestyle and exercise plan by operating a terminal 1201 such as a portable smartphone. For example, in the case of exercise, the subject inputs planned information such as the type of exercise, exercise time, and exercise frequency, and the terminal 1201 creates the plan data 111 shown in FIG. 1.

[0089] The information processing device 100 performs 1. training of a machine learning model, and 2. estimation of behavioral factors using the created machine learning model.

[0090] 1. Training a machine learning model involves the following steps: 1-1. Generate internal features (behavioral features) related to the behavior of each of multiple people based on a plan (behavioral plan) for each of the multiple people. For example, when generating the behavioral features, label the behavior plan. Labeling, for example, associates the exercise load for each task of the planned exercise. 1-2. Using the generated features (behavioral features) and information on the factors behind the behavior of each of multiple people (psychological questionnaire 125), a machine learning model is trained (learning process) to estimate the factors behind the behavior of the target person.

[0091] Explained with reference to FIG. 12, 1. When training a machine learning model, the information processing device 100 acquires plan data 111 and generates internal features (behavioral features) related to the behavior of each person based on the data group of the acquired plan data 111. When generating the behavioral features, the data group of the plan data 111 is labeled. In the labeling, for example, an exercise load (Level) is assigned for each of the exercise tasks described above (step S1201).

[0092] Then, the information processing device 100 estimates the behavioral factors of the subject based on the behavioral features quantified by labeling (step S1202). For example, the behavioral factors to be estimated include self-efficacy, challenge-oriented motivation, and the like.

[0093] 2. When estimating behavioral factors using the created machine learning model, the following process is carried out. The behavioral factor estimation model 106 is trained using internal characteristics (behavioral characteristics) related to the behavior of each of the multiple persons, which are generated based on the plans related to the behavior of each of the multiple persons, and correct answer information related to the factors behind the behavior of each of the multiple persons, and the target person's behavioral plan is inputted to estimate the target person's behavioral factors.

[0094] Then, based on the estimation results of the behavioral factors by the information processing device 100, the administrator can make individual encouragement to each subject. In the example shown in FIG. 12, based on the estimation results of the behavioral factors output by the information processing device 100, the health exercise promotion administrator suggests low-impact exercise if the estimation result of subject A is "low confidence" (in exercise). Also, if the estimation result of subject B is "high spirit of challenge" (in exercise), the administrator encourages the subject with an inspiring message. For example, the administrator transmits these messages to the terminals 1201 such as smartphones of subjects A and B. However, the information processing device 100 may also make individual encouragement to each subject.

[0095] For example, the information processing device 100 may encourage the subject in the following ways: 1. When a subject with low self-efficacy begins to show poor progress, the information processing device 100 suggests changing the exercise plan to make it easier to exercise and to prevent the subject from giving up on exercising. 2. For a subject with high achievement / challenge motivation, the information processing device 100 suggests changing the plan to make the exercise more difficult by increasing the exercise load and frequency, and sends an encouraging message to the subject to prevent the subject from getting bored of exercising. In either case, the information processing device 100 encourages the subject to continue exercising.

[0096] In the embodiment, the system supports behavioral change in a subject by estimating the psychological (internal) characteristics that make it easier for the subject to take the first step and want to continue that behavior, as well as the behavioral factors that are important elements. For example, in response to the social challenges of promoting healthy exercise and extending healthy life expectancy, the system supports behavioral change and the continuation of desirable behaviors, such as daily exercise and a regular lifestyle. Behavioral factors include self-efficacy and intrinsic motivation. What behavioral factors are perceived as high (or low) varies from person to person and also changes over time.

[0097] By changing the approach to suit each individual's different behavioral factors, it is possible to achieve behavioral change that is appropriate for each individual. Behavioral factors can be quantified using a psychological questionnaire consisting of over a dozen questions125, but this is a heavy burden to carry out on a daily basis.

[0098] In the embodiment, behavioral factors are estimated based on behavioral data and sensor data that change over time. Here, according to the embodiment, the training (learning) process prevents a decrease in the accuracy of estimating behavioral factors using a machine learning model, even if the behavioral data includes data on behavior caused by factors other than the person's internal state (behavioral factors).

[0099] Fig. 13 is an explanatory diagram of an encouragement corresponding to the estimated behavioral factors of a subject according to the present invention. Fig. 13 illustrates an example in which the information processing device 100 performs an encouragement process based mainly on the level / low of the estimated behavioral factors.

[0100] The information processing device 100 performs preprocessing (generation of behavioral features, etc.) on a data group of plan data 111 of multiple people in a log data preprocessing unit 101. During learning, a machine learning unit 105 uses the generated behavioral features and a psychological questionnaire 125 to train (learn) a machine learning model that estimates the factors behind the behavior of a subject.

[0101] When estimating the behavioral factors of a subject, planned data 111 regarding the behavior of the subject is input to the behavioral factor estimation model 106 trained by the machine learning unit 105, and the behavioral factor estimation model 106 estimates and outputs whether the behavioral factors of the subject are high or low (step S1301).

[0102] The information processing device 100 performs control in accordance with each behavioral factor. For example, it selects an approach that corresponds to the level (high or low) of the behavioral factor "confidence" (step S1302). For a subject with a low behavioral factor "confidence," it sends a message affirming the current situation (step S1303). Alternatively, as in FIG. 12, it may suggest low-impact exercise to the subject. On the other hand, for a subject with a high behavioral factor "confidence," it sends no message or a standard message that helps the subject maintain high confidence (step S1304).

[0103] Furthermore, the information processing device 100 selects an approach method corresponding to the level of the behavioral factor "spirit of challenge" (step S1305). For subjects with a high behavioral factor "spirit of challenge", an inspiring message is sent to the subject to encourage them (step S1306), as in Fig. 12. On the other hand, for subjects with a low behavioral factor "spirit of challenge", no message is sent, or a standard message that does not discourage the subject's spirit of challenge is sent (step S1307).

[0104] As described above, according to the information processing of the embodiment, a behavioral plan indicating the planned behavior of each of a plurality of individuals is acquired, and behavioral features indicating feature quantities related to the behavior of each of the plurality of individuals are generated based on the acquired behavioral plan. Then, a machine learning model that estimates the behavioral factors of a subject is trained using the feature quantities of the generated behavioral features and correct answer information related to the behavioral factors of each of the plurality of individuals. This allows for the quantification and learning of trends and biases in behavioral load, frequency, number of times, etc. as behavioral features, thereby enabling machine learning to remove noise from the behavioral plan, such as factors unrelated to behavioral factors such as weather, which are attributable to factors other than the individual's internal state, and noise caused by unexpected events, etc.

[0105] Furthermore, according to the information processing of the embodiment, the behavioral plan of the target person's behavior is input to a machine learning model trained using feature quantities of behavioral features indicating feature quantities related to the behavior of each of the multiple people generated based on a behavior plan indicating the plan of the behavior of each of the multiple people and information on correct answers related to the behavioral factors of each of the multiple people, and the behavioral factors of the target person are estimated. This makes it possible to accurately estimate the behavioral factors of the target person without being affected by information attributable to factors other than the person's internal state included in the behavior plan.

[0106] Furthermore, according to the information processing of the embodiment, the generating process may be a process of labeling the load of each behavior included in the acquired behavior plan and generating behavioral features indicating feature amounts related to the behavior of each of multiple people based on the labeled behavior plan. By labeling the load of each behavior included in the behavior plan, the feature amounts of the behavioral features can be quantified, thereby improving the accuracy of machine learning.

[0107] Furthermore, according to the information processing of the embodiment, the generation process may be a process of generating behavioral features based on a behavior plan for a predetermined period and the subject's implementation results for the behavior plan. This makes it possible to process the behavior plan for the predetermined period from among a huge amount of data such as log data, and to estimate behavioral factors at the time when the predetermined period has elapsed.

[0108] According to the information processing of the embodiment, the process of generating the action plan for a predetermined period and the action plan for the execution of the action plan are performed. Based on the subject's implementation results for the action plan, the value of the action characteristics regarding the achievement of the action plan is calculated. This may include a process of calculating the action plan for a predetermined period. It will be possible to estimate the degree of achievement as one of the behavioral factors from the results of actual behavior.

[0109] Furthermore, the information processing according to the embodiment may include a process of correcting feature amounts of behavioral features based on a difference between a behavior plan for a predetermined period and the subject's implementation results for the behavior plan, and the training process may include a process of training a machine learning model using the corrected behavioral features and information on correct answers, thereby improving the learning accuracy of the machine learning model.

[0110] Furthermore, the feature quantities of the behavioral features in the information processing of the embodiment may be feature quantities expressed as vectors based on values ​​included in the behavior plan and values ​​included in the implementation results. The learning accuracy of the machine learning model can be improved by using the values ​​of the behavioral features calculated based on the bias or center of gravity of the feature quantity vectors as inputs to the machine learning.

[0111] Furthermore, according to the information processing of the embodiment, the estimation process may include a process of estimating the level of a behavioral factor of the subject and presenting an appropriate behavioral change to the subject based on the level of the estimated behavioral factor. This makes it possible to present a behavior appropriate for the subject's future or a message encouraging the subject to continue their current behavior, etc., in response to the estimated behavioral factor.

[0112] The estimation of behavioral factors according to the above-described embodiment can be applied to various services that require estimating the factors behind a subject's behavior and taking appropriate action. For example, the invention can be applied not only to the support service for promoting healthy exercise as described above, but also to a support service for encouraging subjects to continue taking e-learning courses within a company. Furthermore, the invention can be applied to services in companies and local governments that support employees' healthy behavior, such as supporting staffing and exercise matching to increase employee satisfaction and productivity, and encouraging exercise and smoking cessation. The invention can also be applied to a support service for specific health guidance provided by industrial public health nurses.

[0113] The information processing method described in this embodiment can be realized by executing a prepared program on a computer such as a PC or a workstation. The information processing program described in this embodiment is recorded on a computer-readable recording medium and is executed by being read from the recording medium by the computer. The information processing program described in this embodiment may also be distributed via a network such as the Internet.

[0114] The following additional notes are provided regarding the above-described embodiment.

[0115] (Appendix 1) Obtaining an action plan showing the plan of action for each of a plurality of persons; generating behavioral features indicating feature amounts related to the behavior of each of the plurality of persons based on the acquired behavior plan; training a machine learning model that estimates the behavioral factors of the subject using the feature amounts of the generated behavioral features and information on correct answers regarding the behavioral factors of each of the plurality of persons; An information processing program that causes a computer to execute a process.

[0116] (Supplementary Note 2) A machine learning model trained using feature quantities of behavioral features indicating feature quantities related to the behavior of each of a plurality of people generated based on an action plan indicating a plan of the behavior of each of the plurality of people and information on correct answers related to the behavioral factors of each of the plurality of people, Enter the action plan for the target person's behavior, Estimating a behavioral factor of the subject; An information processing program that causes a computer to execute a process.

[0117] (Appendix 3) The generating process is Labeling the load of each action included in the acquired action plan; generating behavioral features indicating feature amounts related to the behavior of each of the plurality of persons based on the labeled behavior plan; 2. The information processing program according to claim 1,

[0118] (Supplementary Note 4) The generating process is a process of generating the behavioral features based on the behavior plan for a predetermined period and the subject's implementation results for the behavior plan. 2. The information processing program according to claim 1,

[0119] (Supplementary Note 5) The generating process includes a process of calculating a value of the behavioral feature related to the degree of achievement of the behavior plan based on the behavior plan for a predetermined period and the subject's implementation results for the behavior plan. 2. The information processing program according to claim 1,

[0120] (Supplementary Note 6) The method includes a process of correcting the feature amount of the behavioral feature based on a difference between the behavior plan for a predetermined period and the subject's implementation result for the behavior plan, The training process is a process of training the machine learning model using the corrected behavioral features and the correct answer information. 2. The information processing program according to claim 1,

[0121] (Supplementary Note 7) The feature amount of the behavioral feature is a feature amount expressed as a vector based on a value included in the behavior plan and a value included in the implementation result. 2. The information processing program according to claim 1,

[0122] (Appendix 8) The estimating process estimates the degree of the subject's behavioral factor, a process of suggesting appropriate behavioral changes to the subject based on the degree of the estimated behavioral factors; 3. The information processing program according to claim 2, comprising:

[0123] (Appendix 9) Obtaining an action plan showing the plan of action of each of a plurality of persons; generating behavioral features indicating feature amounts related to the behavior of each of the plurality of persons based on the acquired behavior plan; training a machine learning model that estimates the behavioral factors of the subject using the feature amounts of the generated behavioral features and information on correct answers regarding the behavioral factors of each of the plurality of persons; An information processing method characterized in that the processing is executed by a computer.

[0124] (Supplementary Note 10) A machine learning model trained using feature quantities of behavioral features indicating feature quantities related to the behavior of each of a plurality of people generated based on an action plan indicating a plan of the behavior of each of the plurality of people and information on correct answers related to the behavioral factors of each of the plurality of people, Enter the action plan for the target person's behavior, Estimating a behavioral factor of the subject; An information processing method characterized in that the processing is executed by a computer.

[0125] (Appendix 11) Obtaining an action plan showing the plan of action of each of a plurality of persons; generating behavioral features indicating feature amounts related to the behavior of each of the plurality of persons based on the acquired behavior plan; training a machine learning model that estimates the behavioral factors of the subject using the feature amounts of the generated behavioral features and information on correct answers regarding the behavioral factors of each of the plurality of persons; An information processing device comprising a control unit.

[0126] (Supplementary Note 12) A machine learning model trained using feature quantities of behavioral features indicating feature quantities related to the behavior of each of a plurality of people generated based on an action plan indicating a plan of the behavior of each of the plurality of people and information on correct answers related to the behavioral factors of each of the plurality of people, Enter the action plan for the target person's behavior, Estimating a behavioral factor of the subject; An information processing device comprising a control unit. [Explanation of symbols]

[0127] 100 Information processing device 101 Log data preprocessing unit 102 Vector conversion section 103 Behavioral feature calculation unit 104 Weighting correction unit 105 Machine Learning Department 106 Behavioral Factor Estimation Model 110 Log Data 111 Planning Data 112 Implementation result data 121 Exercise Task List 123 Behavioral Characteristics List 124 Correction Data 125 Psychological Questionnaire 302 Behavioral Data Extraction Layer 306 Survey result storage unit 901 CPU 902 memory 903 Network Interface 905 Recording Media

Claims

1. obtaining an action plan indicating a plan of action for each of a plurality of persons; generating behavioral features indicating feature amounts related to the behavior of each of the plurality of persons based on the acquired behavior plan; training a machine learning model that estimates the behavioral factors of the subject using the feature amounts of the generated behavioral features and information on correct answers regarding the behavioral factors of each of the plurality of persons; An information processing program that causes a computer to execute a process.

2. A machine learning model trained using feature amounts of behavioral features indicating feature amounts related to the behavior of each of a plurality of people generated based on an action plan indicating a plan of the behavior of each of the plurality of people and information on correct answers related to the behavioral factors of each of the plurality of people, Enter the action plan for the target person's behavior, Estimating a behavioral factor of the subject; An information processing program that causes a computer to execute a process.

3. The generating process includes: Labeling the load of each action included in the acquired action plan; generating behavioral features indicating feature amounts related to the behavior of each of the plurality of persons based on the labeled behavior plan; 2. The information processing program according to claim 1, wherein:

4. the generating process is a process of generating the behavioral features based on the behavior plan for a predetermined period and an implementation result of the subject with respect to the behavior plan.

2. The information processing program according to claim 1, wherein:

5. correcting the feature amounts of the behavioral features based on a difference between the behavior plan for a predetermined period and an implementation result of the behavior plan by the subject; The training process is a process of training the machine learning model using the corrected behavioral features and the correct answer information.

2. The information processing program according to claim 1, wherein:

6. The feature amount of the behavioral feature is a feature amount expressed as a vector based on a value included in the behavior plan and a value included in the implementation result.

2. The information processing program according to claim 1, wherein:

7. The estimating process estimates a degree of a behavioral factor of the subject, a process of suggesting appropriate behavioral changes to the subject based on the degree of the estimated behavioral factors; 3. The information processing program according to claim 2, further comprising:

8. obtaining an action plan indicating a plan of action for each of a plurality of persons; generating behavioral features indicating feature amounts related to the behavior of each of the plurality of persons based on the acquired behavior plan; training a machine learning model that estimates the behavioral factors of the subject using the feature amounts of the generated behavioral features and information on correct answers regarding the behavioral factors of each of the plurality of persons; An information processing method characterized in that the processing is executed by a computer.

9. A machine learning model trained using feature amounts of behavioral features indicating feature amounts related to the behavior of each of a plurality of people generated based on an action plan indicating a plan of the behavior of each of the plurality of people and information on correct answers related to the behavioral factors of each of the plurality of people, Enter the action plan for the target person's behavior, Estimating a behavioral factor of the subject; An information processing method characterized in that the processing is executed by a computer.

10. obtaining an action plan indicating a plan of action for each of a plurality of persons; generating behavioral features indicating feature amounts related to the behavior of each of the plurality of persons based on the acquired behavior plan; training a machine learning model that estimates the behavioral factors of the subject using the feature amounts of the generated behavioral features and information on correct answers regarding the behavioral factors of each of the plurality of persons; An information processing device comprising a control unit.

11. A machine learning model trained using feature amounts of behavioral features indicating feature amounts related to the behavior of each of a plurality of people generated based on an action plan indicating a plan of the behavior of each of the plurality of people and information on correct answers related to the behavioral factors of each of the plurality of people, Enter the action plan for the target person's behavior, Estimating a behavioral factor of the subject; An information processing device comprising a control unit.

Images