Policy Information Processing System and Policy Information Processing Method

Abstract

Provided is a policy information processing device capable of discriminating factors contributing to the effectiveness of a policy. 【Solution means】 A policy information processing device including a processor for processing information related to a policy, wherein the processor obtains first policy data related to one or more first policies, environmental data related to the environment in which the first policy was executed, person data related to the people present in the environment at the time of execution of the first policy, and evaluation data related to evaluation indicators of the execution result of the first policy, and obtains first policy result data including the same, and generates a first learning model generated based on at least a part of the first policy data and the first policy result data, and based on the second policy data related to the second policy, derives factor data related to factors for improving the evaluation indicators of the execution result of the second policy and contribution degree data related to the contribution degree of the factor data to the evaluation indicators of the second policy, to generate a first learning model, obtains the second policy data, and based on the second policy data and the first learning model, derives the factor data and the contribution degree data of the second policy.

Description

Technical Field

[0001] The present disclosure relates to a policy information processing system and a policy information processing method.

Background Art

[0002] Conventionally, a data analysis system that analyzes business data to generate measures for improving business operations is known. This data analysis system generates measures for controlling the business operations executed by the business system. The data analysis system includes an arithmetic unit, a storage device connected to the arithmetic unit, and at least one computer connected to the arithmetic unit and having an interface for communicating with the business system. The arithmetic unit acquires, via the interface, business data including a plurality of attributes related to the business from the business system, and based on the distribution of relevant indicators related to the business evaluation indicators for evaluating the business, which are the values of the attributes or the values calculated based on the values of the attributes, identifies the target data to be analyzed, calculates the awareness feature amounts that may contribute to the improvement of the business evaluation indicators by analyzing the target data, generates measures for improving the business evaluation indicators based on the awareness feature amounts, and outputs the data of the generated measures via the interface (see Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] A conventional data analysis system determines whether a measure is effective by using a business that performs control based on the measure data and a business that does not perform control based on the measure data. However, even when a measure is effective, it is difficult to determine what factors cause the measure to be effective.

[0005] The present disclosure provides a policy information processing apparatus and a policy information processing method capable of discriminating factors affecting the effectiveness of policies.

Means for Solving the Problems

[0006] One aspect of the present disclosure is a policy information processing apparatus including a processor that processes information related to policies, wherein the processor obtains first policy data related to one or more first policies, environmental data related to the environment in which the first policies are executed, person data related to persons present in the environment when the first policies are executed, and evaluation data related to evaluation indicators of execution results of the first policies, and includes first policy result data. Use a part of the first policy data as example data, and use the first policy data including the example data, the first policy result data obtained by executing the policy of the first policy data, and the value of the evaluation index as correct answer data. Generate a first learning model by performing machine learning so that the correct answer data can be obtained from the example data. Specify policy conditions for selecting at least a part of the first policy data via an input device to generate second policy data related to a second policy, and based on the first learning model with the policy conditions as input, derive the second policy data and the predicted value of the evaluation index of the execution result of the second policy data It is a policy information processing apparatus.

[0007] One aspect of the present disclosure is a policy information processing method for processing information related to policies, The processor obtaining first policy result data including first policy data related to one or more first policies, environmental data related to the environment in which the first policies are executed, person data related to persons present in the environment when the first policies are executed, and evaluation data related to evaluation indicators of execution results of the first policies. The processor uses a part of the first policy data as example data, and uses the first policy data including the example data, the first policy result data obtained by executing the policy of the first policy data, and the value of the evaluation index as correct answer data. Generate a first learning model by performing machine learning so that the correct answer data can be obtained from the example data. The processor specifies policy conditions for selecting at least a part of the first policy data via an input device to generate second policy data related to a second policy. The processor derives the second policy data and the predicted value of the evaluation index of the execution result of the second policy data based on the first learning model with the policy conditions as input It is a policy information processing method having the above.

Effects of the Invention

[0008] According to the present disclosure, a policy information processing apparatus and a policy information processing method capable of discriminating factors affecting the effectiveness of policies are provided.

Brief Description of the Drawings

[0009]

Figure 1

Figure 2A

Figure 2B

Figure 2C

Figure 2D

Figure 2E

Figure 3A

Figure 3B

Figure 4

Figure 5

Figure 6A

Figure 6B

Figure 7A

Figure 7B

Mode for Carrying Out the Invention

[0010] Hereinafter, embodiments will be described in detail with reference to the drawings as appropriate. However, overly detailed descriptions may be omitted. For example, detailed descriptions of well-known matters and redundant descriptions of substantially the same configurations may be omitted. This is to avoid making the following description unnecessarily redundant and to facilitate understanding by those skilled in the art. Note that the accompanying drawings and the following description are provided to enable those skilled in the art to fully understand the present disclosure, and are not intended to limit the subject matter described in the claims.

[0011] (First Embodiment) <Configuration of the Measure Information System> FIG. 1 is a block diagram showing a configuration example of a measure information processing apparatus 10 according to the first embodiment of the present disclosure.

[0012] The policy information processing apparatus 10 processes information related to policies. A policy is, for example, an attempt to improve some evaluation index (for example, a performance evaluation index (KPI: Key Performance Indicator)). The evaluation index is, for example, the sales amount or the number of visitors at a store that sells products, or other evaluation indexes.

[0013] As a hardware configuration, the policy information processing apparatus 10 includes a processor 11, a memory 12, a communication device 13, an input device 14, and an output device 15. Each hardware configuration is connected to each other via, for example, an internal bus. The policy information processing apparatus 10 is, for example, a PC (Personal Computer), a smartphone, a tablet terminal, or a mobile terminal, etc.

[0014] Note that the policy information processing apparatus 10 may be a system in which the processor 11, the memory 12, the communication device 13, the input device 14, and the output device 15 are configured as individual devices as a processing device, a storage device, a communication device, an input device, and an output device, respectively. This system may be installed in an on-premises environment, a cloud environment, or other environments, and may also be installed in an environment that spans multiple environments.

[0015] The processor 11 may be configured using, for example, a Central Processing Unit (CPU), a Digital Signal Processor (DSP), or a Graphical Processing Unit (GPU). The processor 11 may also be configured using various integrated circuits (for example, Large Scale Integration (LSI) or Field Programmable Gate Array (FPGA)). The processor 11 realizes the functions of various functional units (modules) by executing the programs held in the memory 12. The processor 11 comprehensively controls each part of the system as the policy information processing apparatus 10 and performs various processes.

[0016] For example, the processor 11 controls the execution of measures, analyzes the executed measures, and optimizes the measures to be executed in the future (for example, after the next time). The analysis of the measures is performed based on the data obtained as a result of the measures (also referred to as measure result data). In the analysis of the measures, the behavior and attributes of people existing in the environment where the measures are executed may be analyzed. Note that in the control of the execution of the measures (for example, the control for executing a festival or a sale described later), some processing may be performed within the measure information processing device 10, or various instructions related to the execution of the measures may be given to an external system or an external device.

[0017] Also, for example, the processor 11 specifies the conditions of the measure to be executed (measure conditions), and based on the measure conditions and the measure result data obtained from the past executed measures, generates measure data (also referred to as new measure data) regarding the measure to be executed in the future (also referred to as a new measure), and predicts the evaluation of the generated measure data. For example, it predicts and derives (for example, calculates) the value of the evaluation index that will be obtained when the generated measure data is executed.

[0018] Also, for example, the processor 11 derives data on the factor that improves the evaluation index by the execution of the measure data, and the contribution degree of this factor to the evaluation index, based on the content of the measure and the measure result data, and may analyze the measure. Also, the processor 11 may perform the derivation of the factor and the contribution degree, the analysis of the measure, and the optimization of the measure on the generated new measure data.

[0019] Also, for example, the processor 11 controls to provide the generated measure data (for example, new measure data) and the prediction information of the evaluation index of this measure data. For example, the processor 11 determines the device to which these information are provided. This determination of the destination is also referred to as the optimization of digital media. Digital media widely includes media (media) composed of hardware or software such as a display (signage, display device), a terminal, a predetermined application of the terminal, the WEB, SNS, etc., which can be digitally controlled.

[0020] The memory 12 includes a primary storage device (e.g., Random Access Memory (RAM) or Read Only Memory (ROM)). The memory 12 may include a secondary storage device (e.g., Hard Disk Drive (HDD) or Solid State Drive (SSD)) or a tertiary storage device (e.g., optical disk or SD card), etc. Further, the memory 12 may be an external storage medium and may be detachable from the policy information processing device 10. The memory 12 may be composed of volatile memory elements or non-volatile memory elements. The memory 12 stores various data, information, programs, etc.

[0021] Also, the memory 12 holds, for example, various databases, tables, and information of learning models. The learning models include a learning model M1 and a learning model M2. Although an example is given where there are two learning models, there may be one or three or more.

[0022] The learning model M1 is generated based on policy data regarding policies executed in the past and policy result data including person data regarding the people who obtained the results of the policies and environment data regarding the environment. The learning model M1 derives, for example, factors useful for improving evaluation indicators and their contribution rates in a new policy (new measure). Specifically, the learning model M1 derives factor data regarding factors for improving the evaluation indicators of the execution results of the new policy and contribution degree data regarding the contribution degree of the factor data to the evaluation indicators of the new policy, based on the new policy data regarding the new policy.

[0023] The learning model M2 is also generated based on policy data regarding policies executed in the past and policy result data including person data regarding the people who obtained the results of the policies and environment data regarding the environment. The learning model M2 determines details regarding the new policy based on conditions (policy conditions) for generating the new policy and generates new policy data.

[0024] The communication device 13 communicates various data or information according to a wired or wireless communication method. The communication method by the communication device 13 may include, for example, communication methods such as Local Area Network (LAN), Wide Area Network (WAN), mobile phone network, or power line communication.

[0025] The communication device 13 communicates various data and information with, for example, external devices. The external devices may include, for example, sensor devices, server devices that receive various information, and external terminals and devices that are destinations for providing various information.

[0026] The sensor device is installed, for example, in the environment where the measures are implemented. The sensor device may include, for example, sensors installed in a store or sensors of a terminal held by a person located in the store. The sensor device can detect various events at any timing. The sensor device may detect various events obtained in a state where the measures are not implemented or various events obtained as a result of the implementation of the measures. The sensor device detects, for example, the position, speed, acceleration, temperature, time, or other events of the sensor device. The communication device 13 may acquire data detected by the sensor device from the sensor device. The communication device 13 may acquire information held by the server device from an external server device. The sensor device may include, for example, various sensors, cameras, and microphones.

[0027] The server device may include, for example, a server device related to any business (such as a server for calculating sales) or a server device that holds information related to a person's attributes. The information held by the server device may be acquired as one of the measure result data.

[0028] The input device 14 may include various buttons, keys, a mouse, a keyboard, a touch panel, a microphone, or other input devices. The input device receives inputs such as various data or information. The input device may be operated by a user who uses, for example, the policy information processing device 10. This user may be, for example, a person in charge or an administrator who considers policies for improving evaluation indicators.

[0029] The evaluation indicator is, for example, the customer lifetime value (LTV), but may be other evaluation indicators. LTV is represented, for example, by the following (Formula 1). LTV = number of customers × number of visiting tenants per customer × revisit cycle per customer × purchase unit price per customer ··· (Formula 1)

[0030] The output device 15 is, for example, a liquid crystal display, an organic EL display, or a printer. The output device displays, prints, or outputs various data or information in other ways. The output by the output device may be confirmed by, for example, a user who uses the policy information processing device 10 or other persons.

[0031] Note that the policy information processing device 10 may not have one or both of the input device 14 and the output device 15.

[0032] Next, various functional units executed by the processor 11 and information held by the memory 12 will be described.

[0033] The processor 11 has, as functional units, a totaling unit 111, a totaling and analysis unit 112, an improvement policy processing unit 113, an improvement factor processing unit 114, and an information providing processing unit 115. The memory 12 has an acquired database (DB) 121.

[0034] The acquired database 121 stores, for example, person data related to people, environmental data related to the environment in which the measures are implemented (also referred to as the measure implementation environment), and measure data related to the measures. The person data and the environmental data are one of the measure result data obtained by the implementation of the measures. The acquired database 121 stores data input from the input device 14, data downloaded or communicated from an external server device or a specific external site via the communication device 13, data acquired from the sensor device via the communication device 13, or data based on this data, and the like.

[0035] The measure data is data related to the content of the measures and is data related to measures that have ended or are in progress, etc. The person data is the attributes and statistical information of people measured inside and outside the environment in which the measures are implemented. This person exists within the environment in which the measures are implemented. The person data further includes flow-of-people data and attribute data. The flow-of-people data is data related to the movement (flow) of people within the measure implementation environment at the time of measure implementation. The attribute data is data indicating the characteristics and properties of people. The environmental data is data related to the meteorological situation and other environments at any timing (e.g., year, month, day) in the measure implementation environment.

[0036] The data held in the acquired database 121 is acquired, for example, at the timing when the measures are implemented or at a timing after the measures are implemented. The processor 11 may appropriately add or delete data items and data in the acquired database 121. Note that the data held in the acquired database 121 may also be acquired at any timing when the measures are not implemented, for example, periodically, or irregularly, or sequentially, or at a timing when there is some trigger.

[0037] The aggregating unit 111 generates a data set DS by associating the measure data and the measure result data for each measure based on the measure data and the measure result data held in the acquisition database 121. The aggregating unit 111 may extract data from the acquisition database 121 in time units to generate the data set DS. The data set DS serves as learning data for the learning model M1 or the learning model M2. The data set DS may be held in the memory 12. The aggregating unit 111 may adjust the data formats of the respective data held in the acquisition database 121 to generate each data set DS. The data set DS may be generated by associating at least a part of the measure data and at least a part of the measure result data for each measure and used as learning data. The learning data may include, for example, example data and correct answer data for the example data, as will be described later.

[0038] FIG. 2A is a diagram showing an example of the data held in the acquisition database 121. FIG. 2B is a diagram showing an example of the environmental data. FIG. 2C is a diagram showing an example of the measure data. FIG. 2D is a diagram showing an example of the pedestrian flow data. FIG. 2E is a diagram showing an example of the human attribute data. Note that the data held in the acquisition database 121 is divided for each measure, and the data obtained by associating the measure data and the measure result data is the data set DS. That is, the data shown in FIG. 2A divided for each measure is also an example of the data included in the data set DS.

[0039] As described above, the dataset DS and the acquisition database 121 include policy data, human data, and environmental data. The policy data includes information such as the classification of the policy, the name of the policy, the content of the policy, the location where the policy was implemented, the capacity of the location where the policy was implemented, whether the location where the policy was implemented is indoors or outdoors, the holding period of the policy, the holding time of the policy, the attributes of the target of the policy (e.g., age group, gender), the digital media or external posting for notifying the policy (also referred to as the promotion method, notification method), whether it is a participation type or a non-participation type, and so on. The environmental data includes information such as the timing when the environmental data was acquired (i.e., when the policy was implemented) (e.g., year, month, day, day of the week), a holiday flag indicating whether the timing when the environmental data was acquired is a holiday, the weather and temperature at the timing when the environmental data was acquired, and so on.

[0040] The above-mentioned human data includes flow data and attribute data. The flow data includes information such as the number of new people in the policy implementation environment, the number of revisits to the policy implementation environment, the revisit cycle to the policy implementation environment, the staying time of people staying in the policy implementation environment, the movement route (departure point (From)) of people in the policy implementation environment, the movement route (destination point (To)) of people in the policy implementation environment, the action history of people in the policy implementation environment, and so on. The attribute data includes information such as the age, gender, nationality, place of residence, interest orientation, access history, viewing status, viewing duration, and so on of people existing in the policy implementation environment. The access history indicates whether access has been made to the promotion data (e.g., promotion content for promotion) regarding the provided policy. The viewing status indicates whether the promotion content has been viewed. The viewing duration indicates the number of seconds for which the promotion content has been viewed. The promotion content is, for example, the information-providing content described later.

[0041] The dataset DS is used for the learning of the learning models M1 and M2. The larger the number of data included in the dataset DS (i.e., the number of data records), the more learning data there will be, enabling the learning models M1 and M2 to perform more learning and improving the accuracy of the learning models M1 and M2.

[0042] Returning to FIG. 1, the aggregation analysis unit 112 analyzes each piece of data held in the acquisition database 121. For example, the aggregation analysis unit 112 may perform an analysis (aggregation analysis) on each piece of data based on the policy data and the policy result data. The result of the aggregation analysis may be held in the memory 12. The aggregation analysis unit 112 may output the aggregation analysis result via the output device 15. Also, the aggregation analysis unit 112 may analyze various pieces of data based on the results of the processing by the aggregation unit 111, the improvement policy processing unit 113, or the improvement factor processing unit 114.

[0043] The improvement policy processing unit 113 receives user input via the input device 14 and inputs various policy conditions for generating a new policy (new measure). The policy conditions may be conditions for selecting at least one of the detailed data (for example, the attributes of the target person of the policy, the policy execution location, the policy execution period, the policy holding time, the sales promotion method) included in the policy data included in the acquisition database 121 or the data set DS.

[0044] Based on the input policy conditions, the improvement policy processing unit 113 derives new policy data (new policy data) regarding a new policy (new measure) and prediction data for evaluating this new policy data. The new policy data derived here is data including the details of the new policy data. Also, the above-mentioned policy conditions are information for selecting a part of the policy data, and can also be said to be the outline of the policy data. That is, the improvement policy processing unit 113 may determine the details of the policy data based on the outline of the policy data. Also, the prediction data for evaluating the new policy data is, for example, the predicted value of the evaluation index of the execution result of the new policy data. The content of the new policy data may be the same as the content of the above-mentioned policy data shown in FIG. 2C.

[0045] Note that the user may input, via the input device 14, as policy conditions, policy conditions that can select factor data with a high contribution rate derived by the improvement factor processing unit 114 described later. Thereby, for example, when generating an even newer new policy, it can be expected that a newer new policy that further improves (enhances) the evaluation index will be generated.

[0046] The improvement measure processing unit 113 may derive new measure data and its evaluation prediction data based on the input measure conditions using the learning model M2. The improvement measure processing unit 113 causes the learning model M2 to be learned to obtain a learned model. The learning model M2 is generated, for example, for each measure based on at least a part of the dataset DS, that is, at least a part of the past measure data and at least a part of the past measure result data.

[0047] During the learning of the learning model M2, the improvement measure processing unit 113 uses various datasets DS, that is, a part of the past measure data, as example data, and uses the measure data (i.e., the details of the measure data) including a part of the dataset DS as the example data, and the measure result data obtained by executing the measure of this measure data and the value of the evaluation index as the correct answer data. The measure data used as the correct answer data is, for example, all the data of that measure data in the dataset DS. The value of the evaluation index used as the correct answer data may be included in the measure result data or may be a value derived based on the measure result data. The improvement measure processing unit 113 causes the learning model M2 to be learned so as to derive new measure data that matches or approximates the specified measure conditions and the predicted values of the measure result data and the evaluation index of the new measure data from the example data. That is, the learning model M2 is learned so as to derive new measure data including a part (a part of the past measure data) of the dataset DS that satisfies the specified measure conditions (i.e., matches or approximates the measure conditions). In this case, the improvement measure processing unit 113 may cause the learning to be performed so as to obtain correct answer data having a larger value of the evaluation index as much as possible from the example data. The predicted value of the evaluation index of the new measure data may be included in the predicted value of the measure result data or may be a value derived based on the predicted value of the measure result data. The learning model M2 is learned, for example, by deep learning. By using the learning model M2 learned in this way, highly effective new measure data can be obtained.

[0048] In other words, the improvement measure processing unit 113 uses at least a part of the dataset DS generated based on past measure data, environmental data, human data, etc. accumulated in the acquisition database 121 as learning data to generate a learning model M2 or relearn (update) a machine-learned learning model M2, and stores it in the memory 12. The algorithm of the learning model M2 is not particularly limited, and known algorithms can be used. For example, linear regression, multiple regression analysis, support vector machine, decision tree, random forest, deep learning using a multi-layer neural network, or other machine learning methods can be mentioned.

[0049] The improvement measure processing unit 113 predicts (generates) new measure data and evaluation prediction data for this new measure data according to the measure conditions input when using the learning model M2, in accordance with the learning model M2.

[0050] Based on the new measure data, the improvement factor processing unit 114 derives factor data regarding factors that improve the evaluation index of the new measure data and contribution degree data regarding the contribution degree of the factor data to the evaluation index of the new measure data. The factor data is data (feature data, elements) that are features constituting the measure data. The contribution degree data is indicated by, for example, a contribution rate. In this case, the improvement factor processing unit 114 may use the learning model M1 to derive factor data and contribution degree data for the new measure based on the new measure data. The learning model M1 is generated, for example, for each measure based on at least a part of the dataset DS, that is, at least a part of the measure data and at least a part of the measure result data.

[0051] The improvement factor processing unit 114 may generate a learned learning model M1 by converting the learned learning model M2. In this case, the improvement factor processing unit 114 may generate the learning model M1 by approximating the learning model M2, which is a complex model, with a linear model or the like. The improvement factor processing unit 114 may use LIME (Local Interpretable Model-agnostic Explanations), for example, in the conversion process (conversion process) from the learning model M2 to the learning model M1. When using LIME, the improvement factor processing unit 114 can learn a linear regression model from the output of the learning model M2 when a part of the feature values are randomly replaced with other values, and derive the contribution degree (contribution rate) of the feature values (factor data) from the regression coefficients.

[0052] During or after the implementation of the new measure, the improvement factor processing unit 114 may calculate each factor data of the generated new measure data and its contribution rate according to the learning model M1, and identify the factor data that is effective in the new measure data. Note that the learning model M1 is an example of XAI (Explainable AI (Artificial Intelligence)).

[0053] Note that the improvement factor processing unit 114 may independently learn the learning model M1 to generate and update the learning model M1. That is, the improvement factor processing unit 114 may cause the learning model M1 to learn and use it as a learned model. For example, when learning the learning model M1, for each of various past measures, at least a part of the data set DS, that is, the measure data (for example, measure content, target (age, gender), participation type of the measure, promotion method) and the measure result data (for example, human data, environmental data, KPI values) are used as learning data, and the learning model M1 may be caused to learn the factors that improve the evaluation index among this measure data and the degree of contribution of these factors. The learning method in this case is not particularly limited.

[0054] Based on the contribution rate calculated by the improvement factor processing unit 114, the information providing processing unit 115 controls the distribution of content to terminals with a high contribution rate (for example, mobile terminals such as smartphones held by each person or display devices installed in stores). For example, control is performed to distribute the optimal distribution destination and the optimal content for that distribution destination, such as limiting it to specific terminals existing in a specific area.

[0055] For example, when the factor data with a high contribution rate (for example, the contribution rate is equal to or higher than a predetermined value) is the location where the measure is executed, the information providing processing unit 115 may determine the terminal of the person located at that location as the terminal for information provision. Whether or not it is located at that location may be determined based on, for example, the position information of the terminal detected by the sensor device, the movement route and action history of the pedestrian flow data included in the acquisition database 121, etc. Also, when the factor data with a high contribution rate is the target person for whom the measure in the measure data is executed, the information providing processing unit 115 may determine the terminal of the target person (for example, men and women in their 20s) that matches that target as the terminal for information provision. Whether or not it is a target person that matches that target may be determined based on, for example, the attribute data of the person included in the acquisition database 121.

[0056] The information providing processing unit 115 generates content for information provision (information providing content) based on, for example, the data of the sales promotion method included in the new measure data. For example, when the data of the sales promotion method in the new measure data is WEB (for example, the company's own WEB), the information providing processing unit 115 generates content in a form viewable on the WEB as the information providing content. For example, when the data of the sales promotion method in the new measure data is SNS (for example, the company's own SNS), the information providing processing unit 115 generates content in a form viewable on the SNS as the information providing content. That is, the information providing processing unit 115 distributes (transmits) it in a form that can be output by, for example, a WEB browser or an SNS application. For example, when the sales promotion data in the new measure data is a newspaper, the information providing processing unit 115 generates content in a form viewable in the newspaper as the information providing content.

[0057] Next, a specific example of the aggregation analysis will be described.

[0058] Based on the data held in the acquisition database 121, the generated new measure data, the predicted values of the measure result data of the new measure data, etc., the aggregation analysis unit 112 generates and analyzes, for example, the actual value and the comparison value of the evaluation index for each measure period, and outputs the analysis result via the output device 15. For example, the aggregation analysis unit 112 may generate graphs and tables such as FIG. 3A and FIG. 3B.

[0059] FIG. 3A is a diagram showing the change in the value of the evaluation index (KPI) over time. In FIG. 3A, for example, the horizontal axis shows the value of the month and day as an example of time, and the vertical axis shows the sales amount, the number of new customers, and the number of repeat visitors during a predetermined period (for example, one day or one month) as an example of the KPI. For example, in FIG. 3A, the smaller the value of the KPI, the lower the effect of the measure at that time, and the larger the value of the KPI, the higher the effect of the measure at that time. Therefore, FIG. 3A is a diagram for comparing the effectiveness of the measure in the time series during the execution of the measure.

[0060] FIG. 3B shows the value of the evaluation index (KPI, for example, the weekly average KPI) for each area where the measure is executed, and the increase or decrease in the value of the evaluation index between a predetermined week and the previous week. The increase or decrease in the value of the evaluation index is shown, for example, by the difference or ratio between the value of the evaluation index for a predetermined week and the previous week, and both the difference and the ratio are shown in FIG. 3B. That is, FIG. 3B is a diagram for comparing the effectiveness of the measure at multiple measure execution locations.

[0061] Next, specific examples of the measure conditions and the new measure data will be described.

[0062] FIG. 4 is a diagram showing an example of the measure conditions.

[0063] In FIG. 4, the measure conditions are exemplified as including the area name indicating the name of the area where the measure is executed, the holding date indicating the date when the measure is executed, the holding time zone indicating the time zone when the measure is executed, the scale at which the measure is executed, the venue where the measure is executed (for example, indoor or outdoor), and the sales promotion method of the measure. The measure conditions may not include some of the information in FIG. 4, and may additionally have other information. In FIG. 4, three sales promotion methods are included, but for the measures generated based on the measure conditions including the sales promotion methods, at least one of these three sales promotion methods may be included. Also, the measure conditions may not include the sales promotion method. The measure conditions may be, for example, the data itself including at least a part of the measure data in the dataset DS. Also, the measure conditions may be information in a broader concept including at least a part of the measure data in the dataset DS (for example, a period including a predetermined time or time, a region or area including a predetermined place).

[0064] FIG. 5 is a diagram showing an example of the derived new measure data.

[0065] In FIG. 5, five pieces of new measure data and their evaluation prediction data are shown. Also, in FIG. 5, the new measure data and their evaluation prediction data are shown in ranking order. The ranking order is determined based on the evaluation prediction data.

[0066] The new measure data includes, for example, data on the content of the measure, the target (target) to whom the measure is to be implemented, information indicating whether it is a participation type or a viewing type, and the promotion method. The improvement measure processing unit 113 may specify the number of new measure data created, for example, via the input device 14. The evaluation prediction data is data for predicting KPI, and may include, for example, the absolute value of the KPI. Further, the evaluation prediction data may include, for example, the ratio (increase rate), that is, the relative value of the KPI, of the value of the KPI obtained by implementing the target new measure to the value of the KPI obtained on average by implementing each past measure. In FIG. 5, the ranking is made in descending order of the increase rate of the KPI. According to FIG. 5, the measure information processing apparatus 10 can present an optimal new measure that meets the measure conditions, for example, by simulation using the learning model M2.

[0067] Next, specific examples of the factor data and the contribution degree data, and specific examples of the evaluation prediction data related to the new measure data will be described.

[0068] The improvement factor processing unit 114 derives the contribution rate for each factor data of the new measure based on the new measure data according to the learning model M1. The aggregation analysis unit 112 may generate and analyze a graph or a table in FIG. 6A for the contribution rate for each factor data of the new measure. The aggregation analysis unit 112 and the improvement factor processing unit 114 may perform prediction and factor analysis of the execution result when the new measure of the generated new measure data is executed.

[0069] FIG. 6A is an example of XAI analysis and is a diagram showing an example of the contribution rate for each factor data of the new measure. In FIG. 6A, as the factor data contributing to the improvement of the value of the evaluation index (KPI) of the new measure, in order from the top, day type, measure (festival), measure (special sale), and location are exemplified. For each factor data, the contribution rate is shown as a ratio between 0 and 1. According to FIG. 6A, the measure information processing apparatus 10 can grasp the factors for improving the evaluation index when the new measure is executed by the learning model M1.

[0070] FIG. 6B is a diagram showing an example of evaluation prediction data related to the derived new measure data. The improvement measure processing unit 113 derives (for example, calculates) a predicted value for an evaluation index (step S18). In FIG. 6B, as an example, predicted values for evaluation indexes (KPIs) corresponding to the day type and the measure (festival) as two element data are displayed. According to FIG. 6B, the measure information processing apparatus 10 can grasp, for example, the prediction result of the evaluation index when a new measure is implemented by the learning model M2.

[0071] In FIG. 6B, each numerical value shows, as an example, the ratio (increase rate) of the value of the KPI obtained by executing the target new measure to the value of the KPI obtained on average by executing each past measure (also referred to as a predetermined standard). In FIG. 6B, for example, when the aggregated analysis unit 112 is new measure data for executing a measure on weekdays and not holding a festival (that is, a measure other than a festival), it shows that it is predicted that the KPI will decrease by 15% compared to the predetermined standard. For example, when the aggregated analysis unit 112 is new measure data for executing a measure on weekends or holidays and not holding a festival, it shows that it is predicted that the KPI will increase by 31% compared to the predetermined standard. For example, when the aggregated analysis unit 112 is new measure data for executing a measure on weekdays and holding a festival, it shows that it is predicted that the KPI will increase by 140% compared to the predetermined standard. For example, when the aggregated analysis unit 112 is new measure data for executing a measure on weekends or holidays and holding a festival, it shows that it is predicted that the KPI will increase by 520% compared to the predetermined standard.

[0072] Note that the derivation and analysis of the contribution rate for each factor data may be performed not only for new measures but also for past measures that have been executed or current measures that are being executed. That is, the improvement factor processing unit 114 may derive the contribution rate for each factor data of a predetermined measure based on the predetermined measure data that has been executed or is being executed in the past according to the learning model M1, and use it for the analysis by the aggregated analysis unit 112.

[0073] The aggregation analysis unit 112 may output the analysis results illustrated in FIGS. 6A and 6B via the output device 15. For example, the analysis results may include the information of the aggregation results shown in FIGS. 3A and 3B, the factor data as a feature amount (improvement feature amount) for improving the evaluation index shown in FIGS. 6A and 6B, the information indicating the contribution rate thereof, or other information. Further, the improvement measure processing unit 113 may output information regarding new measure data and the like. The output here may include transmission to an external device via the communication device 13.

[0074] <Operation of the Measure Information Processing Apparatus> Next, details of the processing executed by the measure information processing apparatus 10 will be described. FIGS. 7A and 7B are flowcharts for explaining the processing executed by the measure information processing apparatus 10.

[0075] First, the processor 11 performs processing such as acquisition, collection, and aggregation of environment data regarding the measure execution environment and person data regarding persons existing in the measure execution environment.

[0076] Specifically, the processor 11 acquires measure data input using the input device 14 or downloads or receives data via the communication device 13 to obtain measure data, and registers the measure data in the acquisition database 121 (step S11). There may be one or more pieces of measure data to be registered.

[0077] The processor 11 distributes information regarding the registered measure data, that is, information regarding the measure, via the communication device 13 (step S12). The distribution destination of the information here is a display (display device) in an arbitrary store, an arbitrary terminal, the company's website, the company's SNS, or the like. For example, information regarding a special sale or a sale may be distributed.

[0078] The processor 11 controls to execute the measure (an example of the first measure) of the registered measure data (an example of the first measure data) (step S13).

[0079] The processor 11 acquires data including human data and environmental data during the execution of the policy measures from the sensor device or an external device, and registers it in the acquisition database 121 as policy measure result data (step S14). The policy measure result data is data indicating the effectiveness obtained by the policy measures. While the policy measures are being executed, the processor 11 may continuously and sequentially acquire and accumulate the policy measure result data.

[0080] The aggregation unit 111 generates a data set DS including the policy measure data registered in the acquisition database 121 and the policy measure result data obtained from the result of the execution of the policy measures indicated by this policy measure data (step S15). One data set DS is generated for one policy measure data. In this case, the aggregation unit 111 may generate the data set DS by conforming various data stored in the data set DS to a data format that can be stored in the data set DS as necessary.

[0081] The aggregation and analysis unit 112 performs analysis (aggregation and analysis) such as behavior analysis, attribute analysis, stay analysis, route analysis, revisit analysis, etc. based on the data held in the acquisition database 121 (step S16).

[0082] For example, the aggregation and analysis unit 112 can perform behavior analysis, stay analysis, or route analysis, etc. by tracking the location information of the same person at each time. For example, the aggregation and analysis unit 112 can perform attribute analysis based on the attribute data of people who have purchased some common product. For example, the aggregation and analysis unit 112 can perform revisit analysis of this person based on the acquisition of the location information of the same person at the same store at different timings (for example, different periods).

[0083] Also, the aggregation and analysis unit 112 outputs (for example, displays, prints) the result of the aggregation and analysis (for example, the information obtained in FIGS. 3A and 3B) as the analysis result via the output device 15 (step S16).

[0084] Subsequently, the processor 11 performs processing related to the evaluation and prediction of the policy measures.

[0085] The user inputs the measure conditions for generating new measures via the input device 14. That is, the improvement measure processing unit 113 inputs the measure conditions via the input device 14 (step S17). In this case, the user may check the output (e.g., display) result of the aggregated analysis, and based on the result of the aggregated analysis, input the measure conditions via the input device 14.

[0086] The improvement measure processing unit 113 generates one or more new measure data based on the input measure conditions using, for example, the learning model M2, predicts the result when this new measure data is executed, and derives (e.g., calculates) the predicted value of the evaluation index, for example (step S18). That is, the improvement measure processing unit 113 derives the new measure data and its evaluation prediction data.

[0087] Based on the derived one or more new measure data and its evaluation prediction data, the improvement measure processing unit 113 determines one of the one or more new measure data as the data of the improvement measure to be executed (step S19). For example, the improvement measure processing unit 113 may determine the measure data whose predicted value of the evaluation index as the derived evaluation prediction data is equal to or greater than the threshold th or the measure data with the maximum value as the new measure data to be executed. That is, the measure information processing apparatus 10 can automatically evaluate and analyze the new measure data and determine the improvement measure.

[0088] The improvement measure processing unit 113 outputs the information regarding the new measure data via the output device 15 (step S19). In this case, the information regarding the determined one new measure data and its evaluation prediction data may be output, or the information regarding the derived one or more new measure data and its evaluation prediction data may be output.

[0089] Subsequently, the processor 11 performs the optimal utilization processing of digital media (e.g., display device or terminal).

[0090] Specifically, the information providing processing unit 115 determines an information providing terminal that is a destination for providing content related to the new measure, which is the determined improvement measure, i.e., the new measure, and the content based on the new measure data (step S20). In this case, the information providing processing unit 115 may determine the information providing terminal and the content based on the contribution rate of the factor data in the new measure data.

[0091] For example, based on the data held in the acquisition database 121, if the information providing processing unit 115 can determine that the probability that a person with a specific attribute passes in front of a display device installed at a specific location at a specific time on a specific day of the week is high, the information providing processing unit 115 determines a content distribution destination (information providing terminal) and content that a person with this specific attribute watches attentively. At least one of the specific location, specific day of the week, specific time, and person with a specific attribute here is information related to the new measure data. For example, the specific location may be factor data with a high contribution rate (e.g., location). For example, the specific day of the week may be the same day as the factor data with a high contribution rate (e.g., day of the week). For example, the specific time may be the same time as the factor data with a high contribution rate (e.g., time). For example, the person with a specific attribute is a person related to the factor data with a high contribution rate (e.g., age group). Note that the installation location of the display device may be held in the memory 12, for example. For example, based on information such as the location, movement route, and action history included in the people flow data of the person data included in the data held in the acquisition database 121 and the attribute data, it may be determined whether the probability that a person with a specific attribute passes in front of a display device installed at a specific location at a specific time on a specific day of the week is high. The information providing processing unit 115 determines the terminal of the person with the above-described specific attribute as the information providing terminal. Also, based on the data of the promotion method included in the new measure data, the information providing content is determined.

[0092] The information providing processing unit 115 controls to distribute information providing content to the information providing terminal via, for example, the communication device 13 (step S20). Note that the information providing processing unit 115 may automatically perform the distribution of this information providing content after determining the information providing terminal and the information providing content via, for example, the communication device 13.

[0093] Note that the information providing terminal is, for example, a digital signage, a mobile phone, or a tablet PC, but is not limited thereto. The optimal content is, for example, a promotion video for customer acquisition, disaster information for visitors during a disaster, or traffic information, but is not limited thereto.

[0094] The processor 11 controls to execute the measure (an example of the second measure) of the derived or determined new measure data (an example of the second measure data) (step S21).

[0095] Subsequently, the improvement factor processing unit 114 derives (for example, calculates) one or more factor data of the new measure and its contribution rate according to the learning model M1 based on the new measure data obtained by being derived or determined (step S22). Then, the improvement factor processing unit 114 outputs the factor data and its contribution rate data (for example, the information obtained in FIGS. 6A and 6B) via the output device 15 (step S22). Further, the aggregation analysis unit 112 may compare the factor data and the contribution rate for each of the derived new measure data, or compare the factor data and the contribution rate of the determined new measure data and the measure data regarding the past measures, analyze the comparison results, and output the comparison results and the analysis results via the output device 15.

[0096] One or more factor data of the derived new measures, their contribution rates, comparison results, and analysis results may be referred to, for example, as measure conditions for generating the next new measure data of the new measures. That is, the user can, for example, check the output (e.g., displayed) factor data and its contribution rate data, and taking into account the factor data with a high contribution rate, input the measure conditions for generating the next new measure data via the input device 14. That is, the improvement measure processing unit 113 can input the next measure conditions via the input device 14.

[0097] After the processing of step S22, the processor 11 proceeds to step S14 in FIG. 7A. That is, the processor 11 may acquire the measure result data after the execution of the new measure data, add it to the data set DS, and perform aggregation analysis. That is, the new measure data generated as a future measure is added to the data set DS together with its measure result data as the measure data of the executed measures after the execution of the measure, and is used for the learning of the learning models M1 and M2, that is, added to the data used for the generation of the learning models M1 and M2. Further, the processor 11 may input the measure conditions of new new measure data, generate new new measure data and its evaluation prediction data, and determine the new new measure data into one. When inputting the measure conditions of the new new measure data, the factor data and contribution degree data of the new measure data may be referred to. Further, the processor 11 may control to determine and distribute an information providing terminal and information providing content for notifying the measure of the new new measure data. Then, the processor 11 may control to execute the measure of the new new measure data. Then, the processor 11 may derive the factor data and contribution rate of the new new measure data based on the new measure conditions according to the learning model M1. In this way, the measure information processing apparatus 10 can repeat the derivation of the new measure data and the factor analysis of the new measure data (e.g., derivation of the factor data and contribution degree data) to optimize the generated measures.

[0098] Next, the processing regarding the achievement degree of the evaluation index and the evaluation and analysis of the measures will be supplemented.

[0099] As described above, the improvement measure processing unit 113 obtains new measure data and its evaluation prediction data (predicted values of evaluation indicators) by deriving them according to the learning model M2 based on the measure conditions. Further, the improvement measure processing unit 113 obtains, for example, the target values of the evaluation indicators held in the memory 12. The improvement measure processing unit 113 calculates the achievement degree of the evaluation indicator, for example, by calculating the predicted value / target value based on the obtained predicted value and target value of the evaluation indicator.

[0100] For example, when the evaluation indicator is the customer lifetime value (LTV), the improvement measure processing unit 113 classifies the total number of customers set as the target value of the evaluation indicator into existing customers and new customers, and sets the respective target values for existing customers and new customers. Further, the improvement measure processing unit 113 derives predicted values (predicted numbers) for existing customers and new customers as evaluation prediction data of the new measure data according to the learning model M2. In this case, for example, the data included in the learning data used for learning by the learning model M2 may be divided into data for existing customers and data for new customers, and the data related to existing customers and the data related to new customers may be classified and derived when using the learning model M2. The improvement measure processing unit 113 may calculate the achievement degree of the evaluation indicator with respect to the target value for each of the existing customers and new customers based on each target value and each predicted value.

[0101] Further, the improvement measure processing unit 113 may specify the evaluation indicator via, for example, the input device 14. The evaluation indicator may match the detailed data items (elements) included in the measure data, or may be derived (for example, calculated) based on a plurality of data items. The evaluation indicator may be, for example, LTV, the number of new customers per predetermined time, the number of repeat visits, the sales amount, and the like. The improvement measure processing unit 113 may obtain the predicted value of the evaluation indicator of the new measure obtained by the learning model M2 and the actual value of the evaluation indicator obtained by the implementation of the new measure. The aggregation analysis unit 112 may compare and analyze this predicted value and actual value. The aggregation analysis unit 112 may output information regarding this predicted value and actual value via the output device 15.

[0102] The improvement measure processing unit 113 may plot, in terms of numerical values and / or ratios, where the evaluation value calculated from the predetermined data held in the acquisition database 121 is positioned with respect to the levels of the evaluation indicators, and generate output information. The improvement measure processing unit 113 may output this output information via the output device 15. Thereby, the user can confirm the achievement degree of the evaluation indicators.

[0103] The improvement measure processing unit 113 may attempt to improve new measures, that is, new measure data, according to the achievement degree of the evaluation indicators. For example, the improvement measure processing unit 113 may improve (change) the new measure data so that the achievement degree of the evaluation indicators becomes higher. For example, as described above, by repeating the derivation of new measure data and the derivation of its factor data and contribution degree data, and inputting the measure conditions of new measures in consideration of the factor data and the contribution degree data at this time, the achievement degree of the evaluation indicators can be increased. Further, the improvement measure processing unit 113 may cause the output device 15 to output evaluation prediction data including the predicted value of the evaluation indicator and the predicted value of the achievement degree of the evaluation indicator when the new measure data is executed.

[0104] In addition, the improvement factor processing unit 114 may specify factor data regarding the factors (improvement factors) for improving the evaluation indicators of the new measure data, and calculate the contribution rate of each factor data. In this case, the improvement factor processing unit 114 may derive factor data and contribution rate data based on the new measure data according to the learning model M1, and determine which data contributes in the improvement measures based on the contribution rate. The improvement factor processing unit 114 may output information regarding the derived improvement factors and contribution rates via the output device 15. Thereby, the measure information processing apparatus 10 can assist in explaining the optimality of the improvement measures to, for example, the company or customers.

[0105] The user can check the output of the factor data and contribution rate of the new policy data, and thus grasp the factor data and contribution rate. The user can input policy conditions for generating further new policy data in consideration of the factor data and contribution rate. The improvement policy processing unit 113 may specify policy conditions such as the holding date, holding time zone, scale, and sales promotion media that take into account, for example, the derived factor data or contribution rate, via the input device 14. The improvement policy processing unit 113 may derive new policy data according to the learning model M2 based on the policy conditions.

[0106] Based on the new policy data or the further new policy data, the information providing processing unit 115 may determine the information providing terminal and the information providing content, and control to distribute the information providing content to the information providing terminal. Thereby, the policy information processing apparatus 10 can, for example, increase the awareness of the new policy data or the further new policy data for the owner of the information providing terminal and arouse interest, and it can be easily connected to the purchasing behavior of this owner. Therefore, an effect of improving the cost effectiveness for the distributed content is expected.

[0107] By repeatedly executing the processes of the flows shown in FIGS. 7A and 7B, the policy information processing apparatus 10 continuously accumulates the records held in the acquisition database 121, and the data set DS also continuously increases corresponding to the continuously accumulated records. Therefore, an improvement in the accuracy of the learning models M1 and M2 can be expected. In addition, an improvement in the optimization of the information providing terminal and the information providing content can also be expected for the policy information processing apparatus 10. As a result, when the policy information processing apparatus 10 executes the policy of the new policy data and the further new policy data, that is, by repeatedly executing the flows of FIGS. 7A and 7B to sequentially generate and execute improvement policies, an improvement in the evaluation index, for example, an increase in the probability of improving the profit, can also be expected.

[0108] Next, a comparison is made between the data formats of Patent Document 1 and the present embodiment.

[0109] In Patent Document 1, the data of the explanatory variables used for the target variable is limited to the data stored in the business system. Further, in Patent Document 1, when the scope of the business related to the measure is wide or the target variable varies greatly depending on measures such as age, gender, nationality, geography, and climate for the target variable, it is not possible to convert data with different data formats into the same data format and generate a measure with a high evaluation index based on the correlation and contribution degree of complex factors.

[0110] On the other hand, in the present embodiment, the aggregation unit 111 can convert the formats of various data stored in the acquisition database 121 into the dataset DS. For example, even if the data obtained as the measure result data is data distributed in a predetermined standard, the aggregation unit 111 can automatically convert it into the data format of the dataset DS. Therefore, even if there are various data with different data formats, the aggregation unit 111 can convert them into a dataset DS in a unified format and utilize various data.

[0111] As described above, the measure information processing apparatus 10 of the present embodiment generates the learning model M2 based on the measure data of the measures implemented in the past and the measure result data thereof, and can generate an optimal improvement measure (an example of the second measure) based on the measure conditions using the learning model M2. The measure information processing apparatus 10 can evaluate a new measure (an example of the second measure) by automatically analyzing data regarding one or more factors (for example, complex factors) in the new measure data. For example, in the scenario of attracting customers in an event business, the measure information processing apparatus 10 can analyze the prediction of the evaluation index regarding the event business obtained as a result of the new measure and identify factors contributing to the improvement of the evaluation index. Further, the measure information processing apparatus 10 can take this factor into account and generate a further new improvement measure, for example, for maximizing economic benefits. In addition, the measure information processing apparatus 10 can distribute information provision content in a form in which an improvement in the evaluation index is expected to an information provision terminal related to a person for whom an improvement in the evaluation index is expected for the purpose of notifying the improvement measure.

[0112] With the changes and diversification of the lifestyle of consumers, such as the expansion of the e-commerce market, there is a demand for improving LTV (an example of an evaluation index) in real spaces such as cities and commercial facilities where social life is carried out (maximizing economy / profit). Examples of evaluation indices include an increase in the number of customers, an expansion of the circular / recirculation property, and an expansion of the residence / stay time. The policy information processing apparatus 10 can generate reasonable policies through visualization of information related to the evaluation indices. The policy information processing apparatus 10 acquires changes in the flow of people, the attributes of people, the surrounding environment, etc., such as flow-of-people data, people's attribute data, and environmental data, and can automatically perform deep learning on the learning models M1 and M2 by registering (manually or automatically) policy data and policy result data related to the policies. For example, the policy information processing apparatus 10 can generate learning models M1 and M2 that take into account the flow of people and the attributes of people, and can derive highly effective new policy data and grasp the factor data and contribution degree data of the new policy data. For example, the policy information processing apparatus 10 can generate learning models M1 and M2 that take into account the stay time, movement route, or action history of people, and can derive highly effective new policy data and grasp the factor data and contribution degree data of the new policy data. In this way, the policy information processing apparatus 10 can generate new policies (learning model M2 as predictive AI), analyze new policies (learning model M1 as explainable AI: XAI), predict the next new policy for the purpose of improving the KPI achievement rate (learning model M2 as predictive AI), and perform information distribution using digital media such as signage, WEB, and SNS.

[0113] (Overview of the Embodiment) As described above, the present disclosure describes at least the following matters. In the parentheses, corresponding components, etc. in the above-described embodiment are exemplified, but the present disclosure is not limited thereto.

[0114] (Item 1) A policy information processing apparatus (policy information processing apparatus 10) including a processor (processor 11) and processing information related to policies, wherein the processor Obtain first policy result data including first policy data (executed policies) regarding one or more first policies, environmental data regarding the environment in which the first policies were executed, human data regarding the people present in the environment when the first policies were executed, and evaluation data regarding evaluation indicators of the execution results of the first policies. A first learning model generated based on at least a part of the first policy data and the first policy result data, which, based on second policy data regarding a second policy (new policy), derives factor data regarding factors for improving evaluation indicators of the execution results of the second policy and contribution degree data regarding the contribution degree of the factor data to the evaluation indicators of the second policy. Generate a first learning model (learning model M1). Obtain the second policy data. Based on the second policy data and the first learning model, derive the factor data and the contribution degree data of the second policy. Policy information processing device.

[0115] As a result, the policy information processing device can subdivide and analyze the elements of the second policy data, and grasp the factors with a high contribution degree to the effectiveness of the second policy. Also, when outputting factor data and contribution degree data, analysis regarding factors and contribution degrees can be performed. Further, since the first learning model is generated in consideration of the fact that the first policy result data includes environmental data and human data, the policy information processing device can grasp factor data and contribution degree data in consideration of the state of the environment in which the second policy is executed and the movements of people within the environment. Additionally, by using the first learning model, the policy information processing device can perform large-scale learning, and it is expected that the factors for the effectiveness of the second policy data and their contribution degrees can be specified with high accuracy based on a large number of past achievements.

[0116] (Item 2) The processor For each of the first measures, at least a part of the first measure data and the first measure result data are used as learning data for learning. Based on at least a part of the first measure data, a second measure data including at least a part of the first measure data and evaluation prediction data related to prediction of an evaluation index of an execution result of the second measure data are derived, and a second learning model (learning model M2) is generated. Specify the measure conditions for selecting at least a part of the first measure data and generating the second measure data. Based on the second learning model and the measure conditions, derive the second measure data and the evaluation prediction data of the second measure data. The measure information processing apparatus according to item 1.

[0117] Thereby, the measure information processing apparatus can easily derive the second measure data and its evaluation prediction data by a simple operation of specifying the measure conditions of the second measure desired by the user with reference to, for example, a part or the whole of the first measure data.

[0118] (Item 3) The processor generates the first learning model by approximating the second learning model with a linear model. The measure information processing apparatus according to item 2.

[0119] Thereby, the measure information processing apparatus can easily obtain the first learning model from the second learning model.

[0120] (Item 4) The processor obtains second measure result data including environment data related to the environment in which the second measure was executed, person data related to the person present in the environment at the time of execution of the second measure, and evaluation data related to an evaluation index of the execution result of the second measure, adds the second measure data as the first measure data and the second measure result data as the first measure result data to the data used for generating the first learning model and the second learning model. alternately and repeatedly performing derivation of the second measure data and the evaluation prediction data of the second measure data, and derivation of the factor data and the contribution degree data of the second measure The measure information processing apparatus according to item 2 or 3.

[0121] As a result, the measure information processing apparatus can increase the learning data by adding the obtained second measure data and second measure result data to the learning data set as the first measure data and first measure result data, and can improve the accuracy of the first learning model and the second learning model. Further, when repeatedly deriving the second measure data and the factor data and contribution degree data of the second measure data, for example, by generating new second measure data suitable for factors with high contribution degrees among the second measure data, the effectiveness of the next measure can be improved by taking into account the factors of the effectiveness of the second measure. In addition, when the measure information processing apparatus outputs factor data and contribution degree data, it can assist the user in determining suitable measure conditions for the next measure.

[0122] (Item 5) The human data includes flow data of people regarding the flow of people in the environment and attribute data regarding the attributes of the people. The measure information processing apparatus according to any one of items 1 to 4.

[0123] As a result, the measure information processing apparatus can generate the first learning model and the second learning model taking into account the flow of people and the attributes of people, and can realize prediction of the factor data and contribution degree data of the second measure and generation of the second measure data by taking into account the flow of people and the attributes of people in the environment where the first measure was executed.

[0124] (Item 6) The flow data of people includes data regarding the staying time, movement route, or action history of the people at the location where the first measure was implemented. The measure information processing apparatus according to item 5.

[0125] As a result, the policy information processing device can generate a first learning model that takes into account a person's stay time, movement route, or behavior history, and can predict the factor data and contribution degree data of the second policy and generate the second policy data by taking into account the stay time, movement route, or behavior history of people in the environment where the first policy is executed.

[0126] (Item 7) A plurality of the second policy data are derived, The processor determines, as the second policy data to be executed, the second policy data among the plurality of derived second policy data for which the value of the evaluation prediction data is equal to or greater than a threshold value. The policy information processing device according to any one of Items 2 to 6.

[0127] As a result, even when there are a plurality of suitable second policy data that satisfy the policy conditions, the policy information processing device can adopt the one with a high predicted evaluation as the second policy data, making it easier to improve the evaluation index.

[0128] (Item 8) The second policy data includes target person data regarding the person for whom the second policy data is to be executed and notification method data regarding the notification method of the second policy data. The processor determines a terminal for notifying information regarding the second policy data based on the target person data, determines content for notifying information regarding the second policy data based on the notification method data, and transmits the determined content to the determined terminal via a communication device. The policy information processing device according to any one of Items 2 to 7.

[0129] As a result, the policy information processing device can notify the person suitable for the second policy of the scheduled execution of the second policy by a suitable notification method, making it easier to improve the evaluation index.

[0130] (Item 9) A policy information processing method for processing information related to policies, comprising: obtaining first policy result data including first policy data related to one or more first policies, environment data related to the environment in which the first policies are executed, person data related to persons present in the environment when the first policies are executed, and evaluation data related to evaluation indicators of execution results of the first policies; generating a first learning model generated based on at least a part of the first policy data and the first policy result data, and based on second policy data related to a second policy, deriving factor data related to factors for improving evaluation indicators of execution results of the second policy and contribution degree data related to the contribution degree of the factor data to the evaluation indicators of the second policy; obtaining the second policy data; deriving the factor data and the contribution degree data of the second policy based on the second policy data and the first learning model; A policy information processing method having the above steps.

[0131] Thereby, the policy information processing method can obtain the same effect as in item 1.

[0132] (Item 10) A program for causing a computer to execute the policy information processing method according to item 9.

[0133] Thereby, the program can obtain the same effect as in item 1.

[0134] As described above, various embodiments have been described with reference to the drawings, but it goes without saying that the present disclosure is not limited to such examples. It is obvious that those skilled in the art can conceive of various modification examples or correction examples within the scope described in the claims, and it is naturally understood that they also belong to the technical scope of the present disclosure. Also, within the scope not departing from the gist of the invention, the components in the above embodiments may be arbitrarily combined.

[0135] In addition, the above-described embodiment may also be applicable to a program that realizes the functions of the policy information processing method, which is supplied to a computer (for example, the policy information processing apparatus 10) via a network or various storage media, read and executed by the processor of this computer, and a recording medium storing this program.

Industrial Applicability

[0136] The present disclosure is useful for a policy information processing system, a policy information processing method, etc. that can discriminate factors affecting the effectiveness of policies.

Explanation of Signs

[0137] 10 Policy information processing apparatus 11 Processor 12 Memory 13 Communication device 14 Input device 15 Output device 111 Aggregation unit 112 Aggregation analysis unit 113 Improvement policy processing unit 114 Improvement factor processing unit 115 Information providing processing unit 121 Acquisition database DS Dataset

Claims

1. A policy information processing device that includes a processor and processes information related to a policy, The processor, Acquire first action data relating to one or more first actions, and first action result data including environment data relating to an environment in which the first action was executed, people data relating to people who were present in the environment when the first action was executed, and evaluation data relating to an evaluation index of an execution result of the first action; a part of the first action data is used as example data, the first action data including the example data, the first action result data obtained by implementing the action of the first action data, and the value of the evaluation index are used as correct answer data, and a first learning model is generated by performing machine learning so as to obtain the correct answer data from the example data; selecting at least a portion of the first policy data via an input device and specifying a policy condition for generating second policy data related to a second policy; deriving the second policy data and a predicted value of an evaluation index of an execution result of the second policy data based on the first learning model using the policy conditions as an input; Policy information processing device.

2. The processor, generating a second learning model that derives, based on the second measure data, factor data related to factors that improve the evaluation index of the execution result of the second measure and contribution data related to the contribution of the factor data to the evaluation index of the second measure; The second learning model is generated by a conversion process that approximates the first learning model to a linear regression model; In the conversion process, machine learning of the linear regression model is performed using an output of the first learning model in a case where a part of the factor data that is a feature quantity constituting the second policy data is randomly replaced, and the contribution data of the factor data to the evaluation index of the second policy is derived using a regression coefficient of the linear regression model. deriving the factor data and the contribution data of the second measure based on the second measure data and the second learning model; The policy information processing device according to claim 1 .

3. The processor, acquire second measure result data including environment data relating to an environment in which the second measure was executed, person data relating to people who were in the environment when the second measure was executed, and evaluation data relating to the evaluation index of the execution result of the second measure; Adding the second action data as the first action data and the second action result data as the first action result data to data used for generating the first learning model; alternately and repeatedly executing the derivation of the second policy data and the predicted value of the evaluation index corresponding to the second policy data, and the derivation of the factor data and the contribution degree data of the second policy; The policy information processing device according to claim 2 .

4. The person data includes people flow data related to the flow of people in the environment and attribute data related to attributes of the people. The policy information processing device according to claim 2 .

5. The people flow data includes data regarding the person's stay time, movement path, or behavior history in the location where the first measure is implemented, The policy information processing device according to claim 4.

6. A plurality of the second action data are derived; the processor determines, among the plurality of derived second action data, the second action data in which the predicted value of the evaluation index is equal to or greater than a threshold, as the second action data to be executed; The policy information processing device according to claim 1 .

7. The processor, determining a person present at a policy implementation location where the second policy data is to be implemented based on location information of the terminal detected by the sensor device; When the factor data for which the contribution degree data corresponding to the second policy data is equal to or greater than a predetermined value is the policy implementation location, a terminal of a person present at the policy implementation location is determined as a terminal to which the content is to be delivered. The policy information processing device according to claim 2 .

8. A policy information processing method for processing information related to policies, comprising: A processor acquires first action data related to one or more first actions, and first action result data including environment data related to an environment in which the first action was executed, people data related to people who were present in the environment when the first action was executed, and evaluation data related to an evaluation index of an execution result of the first action; the processor generates a first learning model by performing machine learning so as to obtain the correct answer data from the example data by using a part of the first action data as example data, using the first action data including the example data, and using the first action result data and the value of the evaluation index obtained by executing the action of the first action data as correct answer data; the processor selects, via an input device, at least a portion of the first action data and specifies an action condition for generating second action data related to a second action; The processor derives the second action data and a predicted value of an evaluation index of an execution result of the second action data based on the first learning model using the action condition as an input; The policy information processing method includes the steps of:

9. A program for causing a computer to execute the policy information processing method according to claim 8.

Images