WO2024042627A1 - Information processing device, determination method, and determination program - Google Patents

Information processing device, determination method, and determination program Download PDF

Info

Publication number
WO2024042627A1
WO2024042627A1 PCT/JP2022/031794 JP2022031794W WO2024042627A1 WO 2024042627 A1 WO2024042627 A1 WO 2024042627A1 JP 2022031794 W JP2022031794 W JP 2022031794W WO 2024042627 A1 WO2024042627 A1 WO 2024042627A1
Authority
WO
WIPO (PCT)
Prior art keywords
user
policy
measures
measure
budget
Prior art date
Application number
PCT/JP2022/031794
Other languages
French (fr)
Japanese (ja)
Inventor
大心 伊藤
Original Assignee
三菱電機株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 三菱電機株式会社 filed Critical 三菱電機株式会社
Priority to PCT/JP2022/031794 priority Critical patent/WO2024042627A1/en
Publication of WO2024042627A1 publication Critical patent/WO2024042627A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0207—Discounts or incentives, e.g. coupons or rebates

Definitions

  • the present disclosure relates to an information processing device, a determination method, and a determination program.
  • Measures are being taken to improve profits. For example, a coupon may be provided to the user. Then, the user visits the store and purchases many items while using the coupon. This improves profits. However, if a coupon is provided to a user who plans to visit the store, the price will simply be reduced, and the effectiveness of the measure will be low. Therefore, a technique for providing incentives to users has been proposed (see Patent Document 1).
  • the information processing device of Patent Document 1 provides incentives to users based on the behavior history and background information of a plurality of users.
  • the purpose of this disclosure is to eliminate inequality in measures.
  • the information processing device includes a learned model, behavioral characteristic information indicating behavioral characteristics of each user, attribute information indicating attributes of each user, policy candidate information indicating policy candidates, and values for alleviating inequality of measures.
  • an acquisition unit that acquires budget information indicating a certain measure score and a budget; a generation unit that generates data based on the behavior characteristic information, the attribute information, and the measure candidate information; and a generation unit that generates data based on the generated data and the a prediction unit that predicts the increase in sales or the number of store visits when a measure is implemented based on the learned model; and a prediction unit that performs optimization calculations within the budget using the prediction result and the measure score, and and a decision section that decides on measures.
  • inequality in measures can be resolved.
  • FIG. 1 is a diagram showing a communication system.
  • FIG. 2 is a diagram showing hardware included in an information processing device.
  • FIG. 2 is a block diagram showing the functions of the information processing device. It is a figure showing an example of an action history table.
  • FIG. 3 is a diagram showing an example of a behavioral feature table.
  • FIG. 3 is a diagram showing an example of an attribute table. It is a figure showing an example of a measure result table.
  • FIG. 3 is a diagram showing an example of learning data.
  • 3 is a flowchart illustrating an example of processing executed by a learning section. It is a figure showing an example of a measure candidate table.
  • FIG. 3 is a diagram showing an example of generated data. It is a figure showing an example of a prediction result.
  • FIGS. 1 and (B) are diagrams (Part 1) illustrating an example of a method for calculating a policy score.
  • FIG. 2 is a diagram (part 2) illustrating an example of a method for calculating a policy score.
  • FIG. 3 is a diagram illustrating a specific example of processing executed by a determining unit.
  • FIG. 3 is a diagram showing an example of measures for each user.
  • 3 is a flowchart illustrating an example of processing executed by the information processing device.
  • FIG. 1 is a diagram showing a communication system.
  • the communication system includes an information processing device 100 and a terminal device 200.
  • the information processing device 100 and the terminal device 200 communicate via a network.
  • the information processing device 100 is a device that executes a determination method.
  • the information processing device 100 is a server.
  • the information processing device 100 may be a personal computer, a smartphone, a tablet terminal, or the like.
  • Terminal device 200 is a device used by a user.
  • the terminal device 200 is a smartphone, a tablet terminal, or the like. In FIG. 1, one terminal device is depicted. The number of terminal devices may be two or more.
  • FIG. 2 is a diagram showing hardware included in the information processing device.
  • the information processing device 100 includes a processor 101, a volatile storage device 102, a nonvolatile storage device 103, and a communication interface 104.
  • the processor 101 controls the entire information processing device 100.
  • the processor 101 is a CPU (Central Processing Unit), an FPGA (Field Programmable Gate Array), a DSP (Digital Signal Processor), or the like.
  • Processor 101 may be a multiprocessor.
  • the information processing device 100 may include a processing circuit.
  • the information processing device 100 may include a microcomputer or a System on Chip (SoC).
  • SoC System on Chip
  • the volatile storage device 102 is the main storage device of the information processing device 100.
  • the volatile storage device 102 is a RAM (Random Access Memory).
  • the nonvolatile storage device 103 is an auxiliary storage device of the information processing device 100.
  • the nonvolatile storage device 103 is a ROM (Read Only Memory), an HDD (Hard Disk Drive), or an SSD (Solid State Drive).
  • Communication interface 104 communicates with terminal device 200 .
  • FIG. 3 is a block diagram showing the functions of the information processing device.
  • the information processing device 100 includes a storage section 110, a learning section 120, an acquisition section 130, a generation section 140, a prediction section 150, a determination section 160, and an output section 170.
  • the storage unit 110 may be realized as a storage area secured in the volatile storage device 102 or the nonvolatile storage device 103.
  • a part or all of the learning section 120, the acquisition section 130, the generation section 140, the prediction section 150, the determination section 160, and the output section 170 may be realized by a processing circuit.
  • part or all of the learning unit 120, the acquisition unit 130, the generation unit 140, the prediction unit 150, the determination unit 160, and the output unit 170 may be realized as modules of a program executed by the processor 101.
  • the program executed by processor 101 is also referred to as a determination program.
  • the determination program is recorded on a recording medium.
  • the storage unit 110 may store a behavior history table 111, a behavior feature table 112, an attribute table 113, a policy result table 114, a learned model 115, and a policy candidate table 116.
  • the action history table 111, action feature table 112, attribute table 113, policy result table 114, learned model 115, and policy candidate table 116 will be explained later.
  • the learning unit 120 generates a learned model 115.
  • the functions of the learning section 120 will be explained in detail.
  • the learning unit 120 acquires the behavior history table 111.
  • An action history table 111 is shown.
  • FIG. 4 is a diagram showing an example of an action history table.
  • the behavior history table 111 shows the user's behavior history.
  • the action history table 111 has items such as user ID (identifier), date and time, and stay area.
  • the action history table 111 may also include GPS (Global Positioning System) data and ticket gate entrance/exit history.
  • the learning unit 120 extracts the user's behavior characteristics based on the behavior history table 111.
  • the learning unit 120 registers behavioral features in the behavioral feature table 112.
  • a behavioral feature table 112 is shown.
  • FIG. 5 is a diagram showing an example of a behavioral feature table.
  • the behavior feature table 112 shows behavior characteristics for each user.
  • the behavioral feature table 112 has items such as user ID, average store visit frequency, and average store visit time.
  • the learning unit 120 may extract facilities that the user often uses and times when the user often travels as behavioral features. Further, the learning unit 120 may extract behavior patterns on holidays and weekdays as behavior features.
  • the learning unit 120 acquires the attribute table 113.
  • An attribute table 113 is shown.
  • FIG. 6 is a diagram showing an example of an attribute table.
  • the attribute table 113 shows attributes for each user.
  • the attribute table 113 has items such as user ID, age, gender, and address.
  • the learning unit 120 acquires the policy result table 114.
  • a measure result table 114 is shown.
  • FIG. 7 is a diagram showing an example of a measure result table.
  • the measure result table 114 shows the results of measures taken in the past.
  • the policy result table 114 shows that the user with the user ID "00001" used a 100 yen coupon at the A store, and that the user paid 1500 yen to the A store. Note that 1,500 yen may be expressed as A store sales. Further, the measure result table 114 may include the number of increased visits to the store.
  • the learning unit 120 generates learning data based on the behavior feature table 112, attribute table 113, and policy result table 114. Show training data.
  • FIG. 8 is a diagram showing an example of learning data.
  • Learning data 300 is data generated by learning section 120.
  • the learning unit 120 uses the learning data 300 to generate a trained model 115. Multiple regression analysis may be used to generate the learned model 115. For example, the learning unit 120 performs multiple regression analysis using equation (1).
  • y is the objective variable.
  • the values based on the behavioral feature table 112 and the attribute table 113 are x1 to xi.
  • the values based on the measure result table 114 are z1 to zz.
  • T is measure presence information.
  • ⁇ 1 to ⁇ i and ⁇ 1 to ⁇ j are correction coefficients.
  • ⁇ is a constant term.
  • the policy effect may be calculated using equation (2).
  • a known method may be used to generate the trained model 115.
  • SVM support vector machine
  • GBDT gradient boosting decision tree
  • Meta-Learner methods such as S-Leaner and T-Leaner
  • Causal Tree methods such as Causal Tree and Causal Forest are used. etc. may also be used.
  • the learned model 115 is generated.
  • the trained model 115 is a model that predicts sales or an increase in the number of store visits.
  • the learning unit 120 stores the learned model 115 in the storage unit 110 or an external device connectable to the information processing device 100. Note that the illustration of the external device is omitted.
  • FIG. 9 is a flowchart illustrating an example of processing executed by the learning section.
  • the learning unit 120 extracts the user's behavior characteristics based on the behavior history table 111.
  • the learning unit 120 acquires the behavioral feature table 112, the attribute table 113, and the policy result table 114 from the storage unit 110.
  • the learning unit 120 generates learning data based on the behavioral feature table 112, the attribute table 113, and the policy result table 114.
  • Step S14 The learning unit 120 generates the learned model 115 using the learning data.
  • the learning unit 120 stores the learned model 115 in the storage unit 110.
  • the learned model 115 may be generated by a learning device other than the information processing device 100.
  • the acquisition unit 130 acquires the learned model 115.
  • the acquisition unit 130 acquires the trained model 115 from the storage unit 110 or an external device.
  • the acquisition unit 130 acquires the behavioral feature table 112 and the attribute table 113. For example, the acquisition unit 130 acquires the behavioral feature table 112 and the attribute table 113 from the storage unit 110 or an external device.
  • the behavior feature table 112 is also referred to as behavior feature information.
  • the attribute table 113 is also referred to as attribute information.
  • the acquisition unit 130 also acquires the measure candidate table 116.
  • the acquisition unit 130 acquires the policy candidate table 116 from the storage unit 110 or an external device. A measure candidate table 116 is shown.
  • FIG. 10 is a diagram showing an example of a measure candidate table.
  • the policy candidate table 116 shows policy candidates.
  • the measure candidate table 116 may be expressed as information indicating candidates of measures to be implemented for a plurality of users.
  • the policy candidate table 116 is also referred to as policy candidate information.
  • the measure candidate table 116 in FIG. 10 shows nine measures. For example, policy No. 1 indicates providing a 100 yen coupon for A store. For example, policy No. 2 indicates providing a 100 yen coupon for B store.
  • the generation unit 140 generates data based on the behavior feature table 112, attribute table 113, and policy candidate table 116.
  • the data is input to the learned model 115. Shows the data generated.
  • FIG. 11 is a diagram showing an example of generated data.
  • the generation unit 140 generates nine pieces of data based on the behavioral characteristics, attributes, and nine measures of the user ID "00001."
  • the generation unit 140 similarly generates data corresponding to all users, such as user ID "00002".
  • the prediction unit 150 predicts the increase in sales or the number of store visits when the measures are implemented, based on the data generated by the generation unit 140 and the learned model 115. Specifically, the prediction unit 150 inputs the data to the learned model 115, and the learned model 115 outputs the increase in sales or the number of store visits when the measure is implemented. Show prediction results.
  • FIG. 12 is a diagram showing an example of prediction results.
  • the prediction result in FIG. 12 shows the increase in the number of store visits. For example, a prediction result of "1.2" indicates that if a 100 yen coupon from A store is provided to a user with user ID "00001", the number of visits to the store by the user will increase to "1.2".
  • the acquisition unit 130 acquires a policy score that is a value for alleviating inequality in policies.
  • the acquisition unit 130 acquires the policy score from the storage unit 110 or an external device.
  • the method for calculating the policy score will be explained.
  • FIGS. 13A and 13B are diagrams (part 1) illustrating an example of a method for calculating a measure score.
  • FIG. 13(A) shows policy scores that evaluate whether different users receive the policy on the previous day and today.
  • “Coef 1 ” indicated by the equation in FIG. 13(A) may be set by the user. Furthermore, “Coef 1 ” may be automatically set.
  • “i” in the equation of FIG. 13(A) is set to “0” or “1”. “0” indicates that no measures are taken. “1” indicates that the measure will be taken.
  • the previous day and plan 2 in the table of FIG. 13(A) are referred to. Then, the policy score is calculated using the inner product based on the previous day and Plan 2, and "Coef 1. "
  • FIG. 13(B) shows a measure score that evaluates whether there is no bias toward a specific measure among a plurality of measures.
  • “Coef 2 ” indicated by the equation in FIG. 13(B) may be set by the user. Furthermore, “Coef 2 ” may be automatically set.
  • the case where there is no bias in a specific measure among the plurality of measures is the case of plan 2 in the table of FIG. 13(B).
  • plan 2 indicates that coupon A is provided to 21 people, coupon B is provided to 20 people, and coupon C is provided to 19 people.
  • Plan 2 shows a case where coupons A to C are provided to many people.
  • Plan 2 represents a case where one coupon (for example, Coupon A) is not provided to many people.
  • this case is case 1.
  • the policy score is calculated using the standard deviation based on plan 2 and "Coef 2 ".
  • FIG. 14 is a diagram (part 2) illustrating an example of the method for calculating the policy score.
  • FIG. 14 shows policy scores that evaluate whether the policy is implemented for different users. “Coef 3 ” indicated by the formula in FIG. 14 may be set by the user. Furthermore, “Coef 3 ” may be automatically set. The case where measures are taken for different users is case 2 in the table of FIG. 14. The policy score is calculated using the standard deviation based on plan 2 and "Coef 3. " In this way, the measure score is calculated. The calculation methods shown in FIGS. 13 and 14 are examples. Therefore, the policy score may be calculated by a method other than the above.
  • the acquisition unit 130 acquires budget information indicating the budget. For example, the acquisition unit 130 acquires budget information from the storage unit 110 or an external device.
  • the determining unit 160 uses the prediction results and the policy score to perform optimization calculations within the budget and determines the policy for each user. Specifically, the determining unit 160 performs optimization calculation using equation (3). Further, equation (4) is used as a constraint condition.
  • equation (3) is also called an objective function.
  • the determining unit 160 performs optimization calculation using equation (3) in order to maximize the objective function. Further, in the optimization calculation, an optimization method such as the greedy method, a parameter search method such as gradient descent method, Bayesian optimization, etc. may be used.
  • the processing executed by the determining unit 160 will be explained using a specific example.
  • FIG. 15 is a diagram illustrating a specific example of processing executed by the determining unit.
  • the budget is 600 yen.
  • the determining unit 160 considers measure A within the budget. Measure A is that the user with user ID "00001" is provided with a 300 yen coupon from C store, the user with user ID "00002" is provided with a 100 yen coupon from A store, and the user with user ID "00004" is provided with A store. This indicates that a 200 yen coupon will be provided.
  • the determining unit 160 calculates the total value of the prediction results corresponding to measure A.
  • the determining unit 160 calculates an evaluation value for the measure A using the total value of the prediction results and the measure score.
  • the decision unit 160 considers measure B within the budget.
  • Measure B is that the user with user ID "00001” is provided with a 200 yen coupon from B store, the user with user ID "00002" is provided with a 100 yen coupon from C store, and the user with user ID "00003” is provided with C store. This indicates that a 300 yen coupon will be provided.
  • the determining unit 160 calculates the total value of the prediction results corresponding to measure B.
  • the determining unit 160 calculates an evaluation value for the measure B using the total value of the prediction results and the measure score.
  • the determining unit 160 compares the evaluation value for measure A and the evaluation value for measure B. The determining unit 160 determines that measure A is optimal based on the comparison result. The determining unit 160 repeats the same process and examines the optimal policy. Then, the determining unit 160 determines the optimal policy as a policy for each user. Here, an example of measures for each user will be shown.
  • FIG. 16 is a diagram showing an example of measures for each user.
  • the table in FIG. 16 shows the determined measures for each user. For example, the user with user ID "00001" is provided with a 100 yen coupon from A store as a measure. Further, the table in FIG. 16 shows the prediction results and costs corresponding to the determined measures.
  • the determining unit 160 may calculate a policy score based on the policy selected within the budget, and may use the calculated policy score. For example, the determining unit 160 may calculate a policy score based on policy A, and use the calculated policy score. Furthermore, if the calculated policy score is greater than the policy score acquired by the acquisition unit 130, the determining unit 160 may re-determine the policy.
  • the determining unit 160 may perform optimization calculation using equation (5). Further, equation (4) is used as a constraint condition.
  • the output unit 170 outputs the policy for each user as policy information.
  • FIG. 17 is a flowchart illustrating an example of processing executed by the information processing device.
  • the acquisition unit 130 acquires the trained model 115.
  • the acquisition unit 130 acquires the behavioral feature table 112, the attribute table 113, and the policy candidate table 116.
  • the generation unit 140 generates data based on the behavioral feature table 112, the attribute table 113, and the policy candidate table 116.
  • Step S24 Based on the generated data and the learned model 115, the prediction unit 150 predicts the increase in sales or the number of store visits when the measure is implemented.
  • Step S25 The acquisition unit 130 acquires the policy score.
  • Step S26 The acquisition unit 130 acquires budget information.
  • Step S27 The determining unit 160 uses the prediction result and the policy score to perform optimization calculations within the budget, and determines a policy for each user.
  • Step S28 The output unit 170 outputs the policy for each user as policy information.
  • the information processing device 100 performs optimization calculations in order to maximize the effectiveness of the measures within the budget. Furthermore, when optimization calculations are performed normally, there is a possibility that the calculation result will be output every time that the best coupon is provided to a specific user. For example, the calculation result may be output every time that the user with the user ID "00001" is provided with a 300 yen coupon from A store. In order to prevent such biased calculation results from being output, the information processing device 100 uses the policy score in optimization calculations. Preventing biased calculation results from being output eliminates inequality in measures. Therefore, according to the embodiment, the information processing apparatus 100 can eliminate inequality in measures.
  • Information processing device 101 Processor, 102 Volatile storage device, 103 Non-volatile storage device, 104 Communication interface, 110 Storage unit, 111 Behavior history table, 112 Behavior feature table, 113 Attribute table, 11 4 Measure result table, 115 Learned Model, 116 Measure candidate table, 120 Learning unit, 130 Acquisition unit, 140 Generation unit, 150 Prediction unit, 160 Determination unit, 170 Output unit, 200 Terminal device, 300 Learning data.

Abstract

An information processing device (100) comprises: an acquisition unit (130) that acquires a trained model (115), a behavioral characteristics table (112) indicating behavioral characteristics of each user, an attribute table (113) indicating attributes of each user, a measures candidate table (116) indicating measures candidates, measures scores, which are values for reducing the inequality between measures, and budget information indicating a budget; a generation unit (140) that generates data on the basis of the behavioral characteristics table (112), the attribute table (113), and the measures candidate table (116); a prediction unit (150) that predicts the sales or increase in the number of store visits when measures are implemented, on the basis of the generated data and the trained model (115); and a determination unit (160) that determines measures for each user by performing optimization calculations within the budget using the prediction results and the measures scores.

Description

情報処理装置、決定方法、及び決定プログラムInformation processing device, decision method, and decision program
　本開示は、情報処理装置、決定方法、及び決定プログラムに関する。 The present disclosure relates to an information processing device, a determination method, and a determination program.
　収益を向上させるための施策が行われている。例えば、クーポンが、ユーザに提供される。そして、ユーザは、店舗に来店し、クーポンを使いながら多くの品物を購入する。これにより、収益が向上する。しかし、来店予定のユーザにクーポンが提供された場合、単なる値下げになり、施策の効果が低い。そこで、ユーザにインセンティブを付与する技術が提案されている（特許文献１を参照）。特許文献１の情報処理装置は、複数のユーザの行動履歴及び素性情報に基づいて、ユーザにインセンティブを付与する。 Measures are being taken to improve profits. For example, a coupon may be provided to the user. Then, the user visits the store and purchases many items while using the coupon. This improves profits. However, if a coupon is provided to a user who plans to visit the store, the price will simply be reduced, and the effectiveness of the measure will be low. Therefore, a technique for providing incentives to users has been proposed (see Patent Document 1). The information processing device of Patent Document 1 provides incentives to users based on the behavior history and background information of a plurality of users.
特許６８９９３５０号公報Patent No. 6899350
　ところで、施策によっては、不平等が生じる場合がある。施策の不平等は、サービスに対する不信感を与えると考えられる。 Incidentally, depending on the policy, inequality may arise. Inequality in measures is thought to give rise to distrust in services.
　本開示の目的は、施策の不平等を解消することである。 The purpose of this disclosure is to eliminate inequality in measures.
　本開示の一態様に係る情報処理装置が提供される。情報処理装置は、学習済モデル、ユーザ毎の行動特徴を示す行動特徴情報、前記ユーザ毎の属性を示す属性情報、施策の候補を示す施策候補情報、施策の不平等を緩和するための値である施策スコア、及び予算を示す予算情報を取得する取得部と、前記行動特徴情報、前記属性情報、及び前記施策候補情報に基づいて、データを生成する生成部と、生成された前記データと前記学習済モデルとに基づいて、施策を行ったときの売上又は来店増加回数を予測する予測部と、予測結果と前記施策スコアとを用いて、前記予算内で最適化計算を行い、前記ユーザ毎の施策を決定する決定部と、を有する。 An information processing device according to one aspect of the present disclosure is provided. The information processing device includes a learned model, behavioral characteristic information indicating behavioral characteristics of each user, attribute information indicating attributes of each user, policy candidate information indicating policy candidates, and values for alleviating inequality of measures. an acquisition unit that acquires budget information indicating a certain measure score and a budget; a generation unit that generates data based on the behavior characteristic information, the attribute information, and the measure candidate information; and a generation unit that generates data based on the generated data and the a prediction unit that predicts the increase in sales or the number of store visits when a measure is implemented based on the learned model; and a prediction unit that performs optimization calculations within the budget using the prediction result and the measure score, and and a decision section that decides on measures.
　本開示によれば、施策の不平等を解消することができる。 According to the present disclosure, inequality in measures can be resolved.
通信システムを示す図である。FIG. 1 is a diagram showing a communication system. 情報処理装置が有するハードウェアを示す図である。FIG. 2 is a diagram showing hardware included in an information processing device. 情報処理装置の機能を示すブロック図である。FIG. 2 is a block diagram showing the functions of the information processing device. 行動履歴テーブルの例を示す図である。It is a figure showing an example of an action history table. 行動特徴テーブルの例を示す図である。FIG. 3 is a diagram showing an example of a behavioral feature table. 属性テーブルの例を示す図である。FIG. 3 is a diagram showing an example of an attribute table. 施策結果テーブルの例を示す図である。It is a figure showing an example of a measure result table. 学習データの例を示す図である。FIG. 3 is a diagram showing an example of learning data. 学習部が実行する処理の例を示すフローチャートである。3 is a flowchart illustrating an example of processing executed by a learning section. 施策候補テーブルの例を示す図である。It is a figure showing an example of a measure candidate table. 生成されるデータの例を示す図である。FIG. 3 is a diagram showing an example of generated data. 予測結果の例を示す図である。It is a figure showing an example of a prediction result. （Ａ），（Ｂ）は、施策スコアの算出方法の例を示す図（その１）である。(A) and (B) are diagrams (Part 1) illustrating an example of a method for calculating a policy score. 施策スコアの算出方法の例を示す図（その２）である。FIG. 2 is a diagram (part 2) illustrating an example of a method for calculating a policy score. 決定部が実行する処理の具体例を示す図である。FIG. 3 is a diagram illustrating a specific example of processing executed by a determining unit. ユーザ毎の施策の例を示す図である。FIG. 3 is a diagram showing an example of measures for each user. 情報処理装置が実行する処理の例を示すフローチャートである。3 is a flowchart illustrating an example of processing executed by the information processing device.
　以下、図面を参照しながら実施の形態を説明する。以下の実施の形態は、例にすぎず、本開示の範囲内で種々の変更が可能である。 Hereinafter, embodiments will be described with reference to the drawings. The following embodiments are merely examples, and various modifications can be made within the scope of the present disclosure.
実施の形態．
　図１は、通信システムを示す図である。通信システムは、情報処理装置１００と端末装置２００とを含む。情報処理装置１００と端末装置２００とは、ネットワークを介して、通信する。
　情報処理装置１００は、決定方法を実行する装置である。例えば、情報処理装置１００は、サーバである。また、情報処理装置１００は、パーソナルコンピュータ、スマートフォン、タブレット端末などでもよい。
　端末装置２００は、ユーザが使用する装置である。例えば、端末装置２００は、スマートフォン、タブレット端末などである。図１には、１つの端末装置が描かれている。端末装置の数は、２つ以上でもよい。 Embodiment.
FIG. 1 is a diagram showing a communication system. The communication system includes an information processing device 100 and a terminal device 200. The information processing device 100 and the terminal device 200 communicate via a network.
The information processing device 100 is a device that executes a determination method. For example, the information processing device 100 is a server. Further, the information processing device 100 may be a personal computer, a smartphone, a tablet terminal, or the like.
Terminal device 200 is a device used by a user. For example, the terminal device 200 is a smartphone, a tablet terminal, or the like. In FIG. 1, one terminal device is depicted. The number of terminal devices may be two or more.
　次に、情報処理装置１００が有するハードウェアを説明する。
　図２は、情報処理装置が有するハードウェアを示す図である。情報処理装置１００は、プロセッサ１０１、揮発性記憶装置１０２、不揮発性記憶装置１０３、及び通信インタフェース１０４を有する。 Next, hardware included in the information processing device 100 will be explained.
FIG. 2 is a diagram showing hardware included in the information processing device. The information processing device 100 includes a processor 101, a volatile storage device 102, a nonvolatile storage device 103, and a communication interface 104.
　プロセッサ１０１は、情報処理装置１００全体を制御する。例えば、プロセッサ１０１は、ＣＰＵ（Ｃｅｎｔｒａｌ　Ｐｒｏｃｅｓｓｉｎｇ　Ｕｎｉｔ）、ＦＰＧＡ（Ｆｉｅｌｄ　Ｐｒｏｇｒａｍｍａｂｌｅ　Ｇａｔｅ　Ａｒｒａｙ）、ＤＳＰ（Ｄｉｇｉｔａｌ　Ｓｉｇｎａｌ　Ｐｒｏｃｅｓｓｏｒ）などである。プロセッサ１０１は、マルチプロセッサでもよい。また、情報処理装置１００は、処理回路を有してもよい。さらに、情報処理装置１００は、マイクロコンピュータ、又はＳｏＣ（Ｓｙｓｔｅｍ　ｏｎ　Ｃｈｉｐ）を有してもよい。 The processor 101 controls the entire information processing device 100. For example, the processor 101 is a CPU (Central Processing Unit), an FPGA (Field Programmable Gate Array), a DSP (Digital Signal Processor), or the like. Processor 101 may be a multiprocessor. Further, the information processing device 100 may include a processing circuit. Further, the information processing device 100 may include a microcomputer or a System on Chip (SoC).
　揮発性記憶装置１０２は、情報処理装置１００の主記憶装置である。例えば、揮発性記憶装置１０２は、ＲＡＭ（Ｒａｎｄｏｍ　Ａｃｃｅｓｓ　Ｍｅｍｏｒｙ）である。不揮発性記憶装置１０３は、情報処理装置１００の補助記憶装置である。例えば、不揮発性記憶装置１０３は、ＲＯＭ（Ｒｅａｄ　Ｏｎｌｙ　Ｍｅｍｏｒｙ）、ＨＤＤ（Ｈａｒｄ　Ｄｉｓｋ　Ｄｒｉｖｅ）、又はＳＳＤ（Ｓｏｌｉｄ　Ｓｔａｔｅ　Ｄｒｉｖｅ）である。
　通信インタフェース１０４は、端末装置２００と通信する。 The volatile storage device 102 is the main storage device of the information processing device 100. For example, the volatile storage device 102 is a RAM (Random Access Memory). The nonvolatile storage device 103 is an auxiliary storage device of the information processing device 100. For example, the nonvolatile storage device 103 is a ROM (Read Only Memory), an HDD (Hard Disk Drive), or an SSD (Solid State Drive).
Communication interface 104 communicates with terminal device 200 .
　次に、情報処理装置１００が有する機能を説明する。
　図３は、情報処理装置の機能を示すブロック図である。情報処理装置１００は、記憶部１１０、学習部１２０、取得部１３０、生成部１４０、予測部１５０、決定部１６０、及び出力部１７０を有する。 Next, the functions of the information processing device 100 will be explained.
FIG. 3 is a block diagram showing the functions of the information processing device. The information processing device 100 includes a storage section 110, a learning section 120, an acquisition section 130, a generation section 140, a prediction section 150, a determination section 160, and an output section 170.
　記憶部１１０は、揮発性記憶装置１０２又は不揮発性記憶装置１０３に確保した記憶領域として実現してもよい。
　学習部１２０、取得部１３０、生成部１４０、予測部１５０、決定部１６０、及び出力部１７０の一部又は全部は、処理回路によって実現してもよい。また、学習部１２０、取得部１３０、生成部１４０、予測部１５０、決定部１６０、及び出力部１７０の一部又は全部は、プロセッサ１０１が実行するプログラムのモジュールとして実現してもよい。例えば、プロセッサ１０１が実行するプログラムは、決定プログラムとも言う。例えば、決定プログラムは、記録媒体に記録されている。 The storage unit 110 may be realized as a storage area secured in the volatile storage device 102 or the nonvolatile storage device 103.
A part or all of the learning section 120, the acquisition section 130, the generation section 140, the prediction section 150, the determination section 160, and the output section 170 may be realized by a processing circuit. Furthermore, part or all of the learning unit 120, the acquisition unit 130, the generation unit 140, the prediction unit 150, the determination unit 160, and the output unit 170 may be realized as modules of a program executed by the processor 101. For example, the program executed by processor 101 is also referred to as a determination program. For example, the determination program is recorded on a recording medium.
　記憶部１１０は、行動履歴テーブル１１１、行動特徴テーブル１１２、属性テーブル１１３、施策結果テーブル１１４、学習済モデル１１５、及び施策候補テーブル１１６を記憶してもよい。行動履歴テーブル１１１、行動特徴テーブル１１２、属性テーブル１１３、施策結果テーブル１１４、学習済モデル１１５、及び施策候補テーブル１１６については、後で説明する。 The storage unit 110 may store a behavior history table 111, a behavior feature table 112, an attribute table 113, a policy result table 114, a learned model 115, and a policy candidate table 116. The action history table 111, action feature table 112, attribute table 113, policy result table 114, learned model 115, and policy candidate table 116 will be explained later.
＜学習フェーズ＞
　学習部１２０は、学習済モデル１１５を生成する。学習部１２０の機能を詳細に説明する。
　学習部１２０は、行動履歴テーブル１１１を取得する。行動履歴テーブル１１１を示す。 <Learning phase>
The learning unit 120 generates a learned model 115. The functions of the learning section 120 will be explained in detail.
The learning unit 120 acquires the behavior history table 111. An action history table 111 is shown.
　図４は、行動履歴テーブルの例を示す図である。行動履歴テーブル１１１は、ユーザの行動履歴を示す。行動履歴テーブル１１１は、ユーザＩＤ（ｉｄｅｎｔｉｆｉｅｒ）、日時、及び滞在エリアの項目を有する。また、行動履歴テーブル１１１は、ＧＰＳ（Ｇｌｏｂａｌ　Ｐｏｓｉｔｉｏｎｉｎｇ　Ｓｙｓｔｅｍ）データ、改札入退場履歴を含んでもよい。 FIG. 4 is a diagram showing an example of an action history table. The behavior history table 111 shows the user's behavior history. The action history table 111 has items such as user ID (identifier), date and time, and stay area. The action history table 111 may also include GPS (Global Positioning System) data and ticket gate entrance/exit history.
　学習部１２０は、行動履歴テーブル１１１に基づいて、ユーザの行動特徴を抽出する。学習部１２０は、行動特徴を行動特徴テーブル１１２に登録する。行動特徴テーブル１１２を示す。 The learning unit 120 extracts the user's behavior characteristics based on the behavior history table 111. The learning unit 120 registers behavioral features in the behavioral feature table 112. A behavioral feature table 112 is shown.
　図５は、行動特徴テーブルの例を示す図である。行動特徴テーブル１１２は、ユーザ毎の行動特徴を示す。行動特徴テーブル１１２は、ユーザＩＤ、平均来店頻度、及び平均来店時刻の項目を有する。
　学習部１２０は、ユーザがよく利用する施設、よく移動する時間を行動特徴として、抽出してもよい。また、学習部１２０は、休日及び平日の行動パターンを行動特徴として、抽出してもよい。 FIG. 5 is a diagram showing an example of a behavioral feature table. The behavior feature table 112 shows behavior characteristics for each user. The behavioral feature table 112 has items such as user ID, average store visit frequency, and average store visit time.
The learning unit 120 may extract facilities that the user often uses and times when the user often travels as behavioral features. Further, the learning unit 120 may extract behavior patterns on holidays and weekdays as behavior features.
　学習部１２０は、属性テーブル１１３を取得する。属性テーブル１１３を示す。 The learning unit 120 acquires the attribute table 113. An attribute table 113 is shown.
　図６は、属性テーブルの例を示す図である。属性テーブル１１３は、ユーザ毎の属性を示す。属性テーブル１１３は、ユーザＩＤ、年齢、性別、住所などの項目を有する。 FIG. 6 is a diagram showing an example of an attribute table. The attribute table 113 shows attributes for each user. The attribute table 113 has items such as user ID, age, gender, and address.
　学習部１２０は、施策結果テーブル１１４を取得する。施策結果テーブル１１４を示す。 The learning unit 120 acquires the policy result table 114. A measure result table 114 is shown.
　図７は、施策結果テーブルの例を示す図である。施策結果テーブル１１４は、過去に行われた施策の結果を示す。例えば、施策結果テーブル１１４は、ユーザＩＤ“００００１”のユーザが１００円クーポンをＡストアで使い、かつ当該ユーザがＡストアに１５００円を支払ったことを示す。なお、１５００円は、Ａストアの売上と表現してもよい。また、施策結果テーブル１１４は、来店増加回数を含んでもよい。 FIG. 7 is a diagram showing an example of a measure result table. The measure result table 114 shows the results of measures taken in the past. For example, the policy result table 114 shows that the user with the user ID "00001" used a 100 yen coupon at the A store, and that the user paid 1500 yen to the A store. Note that 1,500 yen may be expressed as A store sales. Further, the measure result table 114 may include the number of increased visits to the store.
　学習部１２０は、行動特徴テーブル１１２、属性テーブル１１３、及び施策結果テーブル１１４に基づいて、学習データを生成する。学習データを示す。 The learning unit 120 generates learning data based on the behavior feature table 112, attribute table 113, and policy result table 114. Show training data.
　図８は、学習データの例を示す図である。学習データ３００は、学習部１２０に生成されたデータである。
　学習部１２０は、学習データ３００を用いて、学習済モデル１１５を生成する。学習済モデル１１５の生成では、重回帰分析が用いられてもよい。例えば、学習部１２０は、式（１）を用いて、重回帰分析を行う。 FIG. 8 is a diagram showing an example of learning data. Learning data 300 is data generated by learning section 120.
The learning unit 120 uses the learning data 300 to generate a trained model 115. Multiple regression analysis may be used to generate the learned model 115. For example, the learning unit 120 performs multiple regression analysis using equation (1).
Figure JPOXMLDOC01-appb-M000001
  
Figure JPOXMLDOC01-appb-M000001
  
　なお、ｙは、目的変数である。行動特徴テーブル１１２及び属性テーブル１１３に基づく値は、ｘ１～ｘｉである。施策結果テーブル１１４に基づく値は、ｚ１～ｚｊである。Ｔは、施策有無情報である。α１～αｉ、及びβ１～βｊは、補正係数である。γは、定数項である。また、施策効果は、式（２）を用いて算出されてもよい。 Note that y is the objective variable. The values based on the behavioral feature table 112 and the attribute table 113 are x1 to xi. The values based on the measure result table 114 are z1 to zz. T is measure presence information. α1 to αi and β1 to βj are correction coefficients. γ is a constant term. Moreover, the policy effect may be calculated using equation (2).
Figure JPOXMLDOC01-appb-M000002
Figure JPOXMLDOC01-appb-M000002
　なお、施策ありの場合とは、Ｔに“１”が入力された場合である。施策なしの場合とは、Ｔに“０”が入力された場合である。 Note that the case with a measure is the case where "1" is input to T. The case of no measures is the case where "0" is input to T.
　また、学習済モデル１１５の生成では、公知の方法が用いられてもよい。例えば、学習済モデル１１５の生成では、サポートベクターマシン（ＳＶＭ）、勾配ブースティング決定木（ＧＢＤＴ）、Ｓ－Ｌｅａｎｅｒ、Ｔ－ＬｅａｎｅｒなどのＭｅｔａ－Ｌｅａｒｎｅｒ手法、Ｃａｕｓａｌ　Ｔｒｅｅ、Ｃａｕｓａｌ　ＦｏｒｅｓｔなどのＣａｕｓａｌ　Ｔｒｅｅ手法などが用いられてもよい。 Additionally, a known method may be used to generate the trained model 115. For example, in generating the trained model 115, support vector machine (SVM), gradient boosting decision tree (GBDT), Meta-Learner methods such as S-Leaner and T-Leaner, and Causal Tree methods such as Causal Tree and Causal Forest are used. etc. may also be used.
　このように、学習済モデル１１５は、生成される。学習済モデル１１５は、売上又は来店増加回数を予測するモデルである。学習部１２０は、記憶部１１０又は情報処理装置１００に接続可能な外部装置に、学習済モデル１１５を格納する。なお、当該外部装置の図は、省略されている。 In this way, the learned model 115 is generated. The trained model 115 is a model that predicts sales or an increase in the number of store visits. The learning unit 120 stores the learned model 115 in the storage unit 110 or an external device connectable to the information processing device 100. Note that the illustration of the external device is omitted.
　次に、学習部１２０が実行する処理を、フローチャートを用いて説明する。
　図９は、学習部が実行する処理の例を示すフローチャートである。
　（ステップＳ１１）学習部１２０は、行動履歴テーブル１１１に基づいて、ユーザの行動特徴を抽出する。
　（ステップＳ１２）学習部１２０は、行動特徴テーブル１１２、属性テーブル１１３、及び施策結果テーブル１１４を記憶部１１０から取得する。
　（ステップＳ１３）学習部１２０は、行動特徴テーブル１１２、属性テーブル１１３、及び施策結果テーブル１１４に基づいて、学習データを生成する。
　（ステップＳ１４）学習部１２０は、学習データを用いて、学習済モデル１１５を生成する。
　（ステップＳ１５）学習部１２０は、学習済モデル１１５を記憶部１１０に格納する。 Next, the process executed by the learning unit 120 will be explained using a flowchart.
FIG. 9 is a flowchart illustrating an example of processing executed by the learning section.
(Step S11) The learning unit 120 extracts the user's behavior characteristics based on the behavior history table 111.
(Step S12) The learning unit 120 acquires the behavioral feature table 112, the attribute table 113, and the policy result table 114 from the storage unit 110.
(Step S13) The learning unit 120 generates learning data based on the behavioral feature table 112, the attribute table 113, and the policy result table 114.
(Step S14) The learning unit 120 generates the learned model 115 using the learning data.
(Step S15) The learning unit 120 stores the learned model 115 in the storage unit 110.
　上記の説明では、情報処理装置１００が学習済モデル１１５を生成する場合を説明した。学習済モデル１１５は、情報処理装置１００以外の学習装置が生成してもよい。 In the above description, the case where the information processing device 100 generates the trained model 115 has been described. The learned model 115 may be generated by a learning device other than the information processing device 100.
＜活用フェーズ＞
　図３に戻って、取得部１３０などの機能を説明する。
　取得部１３０は、学習済モデル１１５を取得する。例えば、取得部１３０は、記憶部１１０又は外部装置から学習済モデル１１５を取得する。 <Utilization phase>
Returning to FIG. 3, the functions of the acquisition unit 130 and the like will be explained.
The acquisition unit 130 acquires the learned model 115. For example, the acquisition unit 130 acquires the trained model 115 from the storage unit 110 or an external device.
　取得部１３０は、行動特徴テーブル１１２及び属性テーブル１１３を取得する。例えば、取得部１３０は、行動特徴テーブル１１２及び属性テーブル１１３を記憶部１１０又は外部装置から取得する。ここで、行動特徴テーブル１１２は、行動特徴情報とも言う。属性テーブル１１３は、属性情報とも言う。
　また、取得部１３０は、施策候補テーブル１１６を取得する。例えば、取得部１３０は、施策候補テーブル１１６を記憶部１１０又は外部装置から取得する。施策候補テーブル１１６を示す。 The acquisition unit 130 acquires the behavioral feature table 112 and the attribute table 113. For example, the acquisition unit 130 acquires the behavioral feature table 112 and the attribute table 113 from the storage unit 110 or an external device. Here, the behavior feature table 112 is also referred to as behavior feature information. The attribute table 113 is also referred to as attribute information.
The acquisition unit 130 also acquires the measure candidate table 116. For example, the acquisition unit 130 acquires the policy candidate table 116 from the storage unit 110 or an external device. A measure candidate table 116 is shown.
　図１０は、施策候補テーブルの例を示す図である。施策候補テーブル１１６は、施策の候補を示す。施策候補テーブル１１６は、複数のユーザに行う施策の候補を示す情報と表現してもよい。施策候補テーブル１１６は、施策候補情報とも言う。
　図１０の施策候補テーブル１１６は、９つの施策を示している。例えば、施策Ｎｏ.１は、Ａストアの１００円クーポンを提供することを示す。例えば、施策Ｎｏ.２は、Ｂストアの１００円クーポンを提供することを示す。 FIG. 10 is a diagram showing an example of a measure candidate table. The policy candidate table 116 shows policy candidates. The measure candidate table 116 may be expressed as information indicating candidates of measures to be implemented for a plurality of users. The policy candidate table 116 is also referred to as policy candidate information.
The measure candidate table 116 in FIG. 10 shows nine measures. For example, policy No. 1 indicates providing a 100 yen coupon for A store. For example, policy No. 2 indicates providing a 100 yen coupon for B store.
　生成部１４０は、行動特徴テーブル１１２、属性テーブル１１３、及び施策候補テーブル１１６に基づいて、データを生成する。当該データは、学習済モデル１１５に入力されるデータである。生成されるデータを示す。 The generation unit 140 generates data based on the behavior feature table 112, attribute table 113, and policy candidate table 116. The data is input to the learned model 115. Shows the data generated.
　図１１は、生成されるデータの例を示す図である。生成部１４０は、ユーザＩＤ“００００１”の行動特徴、属性、及び９つの施策に基づいて、９つのデータを生成する。生成部１４０は、同様に、ユーザＩＤ“００００２”などの全てのユーザに対応するデータを生成する。 FIG. 11 is a diagram showing an example of generated data. The generation unit 140 generates nine pieces of data based on the behavioral characteristics, attributes, and nine measures of the user ID "00001." The generation unit 140 similarly generates data corresponding to all users, such as user ID "00002".
　予測部１５０は、生成部１４０により生成されたデータと学習済モデル１１５とに基づいて、施策を行ったときの売上又は来店増加回数を予測する。詳細には、予測部１５０が当該データを学習済モデル１１５に入力することで、学習済モデル１１５は、施策を行ったときの売上又は来店増加回数を出力する。予測結果を示す。 The prediction unit 150 predicts the increase in sales or the number of store visits when the measures are implemented, based on the data generated by the generation unit 140 and the learned model 115. Specifically, the prediction unit 150 inputs the data to the learned model 115, and the learned model 115 outputs the increase in sales or the number of store visits when the measure is implemented. Show prediction results.
　図１２は、予測結果の例を示す図である。図１２の予測結果は、来店増加回数を示す。例えば、予測結果“１．２”は、ユーザＩＤ“００００１”のユーザにＡストアの１００円クーポンを提供した場合、当該ユーザの来店回数が“１．２”回に増えることを示す。 FIG. 12 is a diagram showing an example of prediction results. The prediction result in FIG. 12 shows the increase in the number of store visits. For example, a prediction result of "1.2" indicates that if a 100 yen coupon from A store is provided to a user with user ID "00001", the number of visits to the store by the user will increase to "1.2".
　取得部１３０は、施策の不平等を緩和するための値である施策スコアを取得する。例えば、取得部１３０は、記憶部１１０又は外部装置から施策スコアを取得する。ここで、施策スコアの算出方法を説明する。 The acquisition unit 130 acquires a policy score that is a value for alleviating inequality in policies. For example, the acquisition unit 130 acquires the policy score from the storage unit 110 or an external device. Here, the method for calculating the policy score will be explained.
　図１３（Ａ），（Ｂ）は、施策スコアの算出方法の例を示す図（その１）である。図１３（Ａ）は、前日と今日で異なるユーザが施策を受けることを評価する施策スコアを示す。図１３（Ａ）の式が示す“Ｃｏｅｆ１”は、ユーザが設定してもよい。また、“Ｃｏｅｆ１”は、自動で設定されてもよい。図１３（Ａ）の式の中の“ｉ”には、“０”又は“１”が設定される。“０”は、施策を受けないことを示す。“１”は、施策を受けることを示す。前日と今日で異なるユーザが施策を受けることを評価する施策スコアを算出する場合、図１３（Ａ）の表における前日と案２とが参照される。そして、施策スコアは、前日と案２とに基づく内積と、“Ｃｏｅｆ１”とを用いて、算出される。 FIGS. 13A and 13B are diagrams (part 1) illustrating an example of a method for calculating a measure score. FIG. 13(A) shows policy scores that evaluate whether different users receive the policy on the previous day and today. “Coef 1 ” indicated by the equation in FIG. 13(A) may be set by the user. Furthermore, “Coef 1 ” may be automatically set. “i” in the equation of FIG. 13(A) is set to “0” or “1”. “0” indicates that no measures are taken. “1” indicates that the measure will be taken. When calculating a policy score that evaluates whether different users receive the policy on the previous day and today, the previous day and plan 2 in the table of FIG. 13(A) are referred to. Then, the policy score is calculated using the inner product based on the previous day and Plan 2, and "Coef 1. "
　図１３（Ｂ）は、複数の施策の中で、特定の施策に偏りがないことを評価する施策スコアを示す。図１３（Ｂ）の式が示す“Ｃｏｅｆ２”は、ユーザが設定してもよい。また、“Ｃｏｅｆ２”は、自動で設定されてもよい。複数の施策の中で、特定の施策に偏りがない場合とは、図１３（Ｂ）の表における案２の場合である。例えば、案２は、Ａクーポンを２１人に提供し、Ｂクーポンを２０人に提供し、Ｃクーポンを１９人に提供することを示す。このように、案２は、Ａ～Ｃクーポンが多くの人に提供される場合を示す。言い換えれば、案２は、１つのクーポン（例えば、Ａクーポン）が多く人に提供されない場合を示している。例えば、当該場合は、案１の場合である。施策スコアは、案２に基づく標準偏差と、“Ｃｏｅｆ２”とを用いて、算出される。 FIG. 13(B) shows a measure score that evaluates whether there is no bias toward a specific measure among a plurality of measures. “Coef 2 ” indicated by the equation in FIG. 13(B) may be set by the user. Furthermore, “Coef 2 ” may be automatically set. The case where there is no bias in a specific measure among the plurality of measures is the case of plan 2 in the table of FIG. 13(B). For example, plan 2 indicates that coupon A is provided to 21 people, coupon B is provided to 20 people, and coupon C is provided to 19 people. In this way, Plan 2 shows a case where coupons A to C are provided to many people. In other words, Plan 2 represents a case where one coupon (for example, Coupon A) is not provided to many people. For example, this case is case 1. The policy score is calculated using the standard deviation based on plan 2 and "Coef 2 ".
　図１４は、施策スコアの算出方法の例を示す図（その２）である。図１４は、異なるユーザにも施策を行うことを評価する施策スコアを示す。図１４の式が示す“Ｃｏｅｆ３”は、ユーザが設定してもよい。また、“Ｃｏｅｆ３”は、自動で設定されてもよい。異なるユーザにも施策を行う場合とは、図１４の表における案２の場合である。施策スコアは、案２に基づく標準偏差と、“Ｃｏｅｆ３”とを用いて、算出される。
　このように、施策スコアは、算出される。図１３，１４の算出方法は、一例である。そのため、施策スコアは、上記以外の方法で算出されてもよい。 FIG. 14 is a diagram (part 2) illustrating an example of the method for calculating the policy score. FIG. 14 shows policy scores that evaluate whether the policy is implemented for different users. “Coef 3 ” indicated by the formula in FIG. 14 may be set by the user. Furthermore, “Coef 3 ” may be automatically set. The case where measures are taken for different users is case 2 in the table of FIG. 14. The policy score is calculated using the standard deviation based on plan 2 and "Coef 3. "
In this way, the measure score is calculated. The calculation methods shown in FIGS. 13 and 14 are examples. Therefore, the policy score may be calculated by a method other than the above.
　また、取得部１３０は、予算を示す予算情報を取得する。例えば、取得部１３０は、記憶部１１０又は外部装置から予算情報を取得する。 Additionally, the acquisition unit 130 acquires budget information indicating the budget. For example, the acquisition unit 130 acquires budget information from the storage unit 110 or an external device.
　決定部１６０は、予測結果と施策スコアとを用いて、予算内で最適化計算を行い、ユーザ毎の施策を決定する。詳細には、決定部１６０は、式（３）を用いて、最適化計算を行う。また、制約条件として、式（４）が用いられる。 The determining unit 160 uses the prediction results and the policy score to perform optimization calculations within the budget and determines the policy for each user. Specifically, the determining unit 160 performs optimization calculation using equation (3). Further, equation (4) is used as a constraint condition.
Figure JPOXMLDOC01-appb-M000003
Figure JPOXMLDOC01-appb-M000003
Figure JPOXMLDOC01-appb-M000004
Figure JPOXMLDOC01-appb-M000004
　ここで、式（３）は、目的関数とも言う。決定部１６０は、目的関数を最大化するために、式（３）を用いて、最適化計算を行う。また、最適化計算では、グリーディ法などの最適化手法、勾配降下法、ベイズ最適化などのパラメータ探索手法が用いられてもよい。 Here, equation (3) is also called an objective function. The determining unit 160 performs optimization calculation using equation (3) in order to maximize the objective function. Further, in the optimization calculation, an optimization method such as the greedy method, a parameter search method such as gradient descent method, Bayesian optimization, etc. may be used.
　決定部１６０が実行する処理を、具体例を用いて説明する。 The processing executed by the determining unit 160 will be explained using a specific example.
　図１５は、決定部が実行する処理の具体例を示す図である。まず、予算は、６００円とする。決定部１６０は、予算内で施策Ａを検討する。施策Ａは、ユーザＩＤ“００００１”のユーザにＣストアの３００円クーポンが提供され、ユーザＩＤ“００００２”のユーザにＡストアの１００円クーポンが提供され、ユーザＩＤ“００００４”のユーザにＡストアの２００円クーポンが提供されることを示す。決定部１６０は、施策Ａに対応する予測結果の合計値を算出する。決定部１６０は、予測結果の合計値と施策スコアとを用いて、施策Ａに対する評価値を算出する。 FIG. 15 is a diagram illustrating a specific example of processing executed by the determining unit. First, the budget is 600 yen. The determining unit 160 considers measure A within the budget. Measure A is that the user with user ID "00001" is provided with a 300 yen coupon from C store, the user with user ID "00002" is provided with a 100 yen coupon from A store, and the user with user ID "00004" is provided with A store. This indicates that a 200 yen coupon will be provided. The determining unit 160 calculates the total value of the prediction results corresponding to measure A. The determining unit 160 calculates an evaluation value for the measure A using the total value of the prediction results and the measure score.
　決定部１６０は、予算内で施策Ｂを検討する。施策Ｂは、ユーザＩＤ“００００１”のユーザにＢストアの２００円クーポンが提供され、ユーザＩＤ“００００２”のユーザにＣストアの１００円クーポンが提供され、ユーザＩＤ“００００３”のユーザにＣストアの３００円クーポンが提供されることを示す。決定部１６０は、施策Ｂに対応する予測結果の合計値を算出する。決定部１６０は、予測結果の合計値と施策スコアとを用いて、施策Ｂに対する評価値を算出する。 The decision unit 160 considers measure B within the budget. Measure B is that the user with user ID "00001" is provided with a 200 yen coupon from B store, the user with user ID "00002" is provided with a 100 yen coupon from C store, and the user with user ID "00003" is provided with C store. This indicates that a 300 yen coupon will be provided. The determining unit 160 calculates the total value of the prediction results corresponding to measure B. The determining unit 160 calculates an evaluation value for the measure B using the total value of the prediction results and the measure score.
　決定部１６０は、施策Ａに対する評価値と施策Ｂに対する評価値とを比較する。決定部１６０は、比較結果に基づいて、施策Ａが最適であると判定する。決定部１６０は、同様の処理を繰り返し、最適な施策の検討を行う。そして、決定部１６０は、最適な施策を、ユーザ毎の施策として決定する。ここで、ユーザ毎の施策の例を示す。 The determining unit 160 compares the evaluation value for measure A and the evaluation value for measure B. The determining unit 160 determines that measure A is optimal based on the comparison result. The determining unit 160 repeats the same process and examines the optimal policy. Then, the determining unit 160 determines the optimal policy as a policy for each user. Here, an example of measures for each user will be shown.
　図１６は、ユーザ毎の施策の例を示す図である。図１６の表は、決定されたユーザ毎の施策を示している。例えば、ユーザＩＤ“００００１”のユーザには、施策として、Ａストアの１００円クーポンが提供される。
　また、図１６の表は、決定された施策に対応する予測結果とコストを示している。 FIG. 16 is a diagram showing an example of measures for each user. The table in FIG. 16 shows the determined measures for each user. For example, the user with user ID "00001" is provided with a 100 yen coupon from A store as a measure.
Further, the table in FIG. 16 shows the prediction results and costs corresponding to the determined measures.
　なお、上記では、取得部１３０が取得した施策スコアを用いる場合を説明した。決定部１６０は、予算内で選択された施策に基づいて、施策スコアを算出し、算出された施策スコアを用いてもよい。例えば、決定部１６０は、施策Ａに基づいて、施策スコアを算出し、算出された施策スコアを用いてもよい。また、決定部１６０は、算出された施策スコアが、取得部１３０により取得された施策スコアよりも大きい場合、施策の決定をやり直してもよい。 Note that the above describes a case where the measure score acquired by the acquisition unit 130 is used. The determining unit 160 may calculate a policy score based on the policy selected within the budget, and may use the calculated policy score. For example, the determining unit 160 may calculate a policy score based on policy A, and use the calculated policy score. Furthermore, if the calculated policy score is greater than the policy score acquired by the acquisition unit 130, the determining unit 160 may re-determine the policy.
　また、決定部１６０は、式（５）を用いて、最適化計算を行ってもよい。また、制約条件として、式（４）が用いられる。 Additionally, the determining unit 160 may perform optimization calculation using equation (5). Further, equation (4) is used as a constraint condition.
Figure JPOXMLDOC01-appb-M000005
Figure JPOXMLDOC01-appb-M000005
　出力部１７０は、ユーザ毎の施策を、施策情報として出力する。 The output unit 170 outputs the policy for each user as policy information.
　次に、情報処理装置１００が実行する処理を、フローチャートを用いて、説明する。
　図１７は、情報処理装置が実行する処理の例を示すフローチャートである。
　（ステップＳ２１）取得部１３０は、学習済モデル１１５を取得する。
　（ステップＳ２２）取得部１３０は、行動特徴テーブル１１２、属性テーブル１１３、及び施策候補テーブル１１６を取得する。
　（ステップＳ２３）生成部１４０は、行動特徴テーブル１１２、属性テーブル１１３、及び施策候補テーブル１１６に基づいて、データを生成する。 Next, the processing executed by the information processing apparatus 100 will be explained using a flowchart.
FIG. 17 is a flowchart illustrating an example of processing executed by the information processing device.
(Step S21) The acquisition unit 130 acquires the trained model 115.
(Step S22) The acquisition unit 130 acquires the behavioral feature table 112, the attribute table 113, and the policy candidate table 116.
(Step S23) The generation unit 140 generates data based on the behavioral feature table 112, the attribute table 113, and the policy candidate table 116.
　（ステップＳ２４）予測部１５０は、生成されたデータと学習済モデル１１５とに基づいて、施策を行ったときの売上又は来店増加回数を予測する。
　（ステップＳ２５）取得部１３０は、施策スコアを取得する。
　（ステップＳ２６）取得部１３０は、予算情報を取得する。
　（ステップＳ２７）決定部１６０は、予測結果と施策スコアとを用いて、予算内で最適化計算を行い、ユーザ毎の施策を決定する。
　（ステップＳ２８）出力部１７０は、ユーザ毎の施策を、施策情報として出力する。 (Step S24) Based on the generated data and the learned model 115, the prediction unit 150 predicts the increase in sales or the number of store visits when the measure is implemented.
(Step S25) The acquisition unit 130 acquires the policy score.
(Step S26) The acquisition unit 130 acquires budget information.
(Step S27) The determining unit 160 uses the prediction result and the policy score to perform optimization calculations within the budget, and determines a policy for each user.
(Step S28) The output unit 170 outputs the policy for each user as policy information.
　ここで、予算が無限であれば、全てのユーザに一番良いクーポンを提供できる。しかし、予算には、制限がある。そのため、情報処理装置１００は、予算内で施策の効果を最大化するために、最適化計算を行う。また、普通に最適化計算が行われた場合、特定のユーザに一番良いクーポンが提供される、という計算結果が毎回出力される可能性がある。例えば、ユーザＩＤ“００００１”のユーザにＡストアの３００円クーポンが提供される、という計算結果が毎回出力される可能性がある。このような偏った計算結果が出力されることを防止するために、情報処理装置１００は、最適化計算で、施策スコアを用いる。そして、偏った計算結果が出力されることを防止することは、施策の不平等を解消する。よって、実施の形態によれば、情報処理装置１００は、施策の不平等を解消することができる。 Here, if the budget is unlimited, the best coupon can be provided to all users. However, there are budget limitations. Therefore, the information processing device 100 performs optimization calculations in order to maximize the effectiveness of the measures within the budget. Furthermore, when optimization calculations are performed normally, there is a possibility that the calculation result will be output every time that the best coupon is provided to a specific user. For example, the calculation result may be output every time that the user with the user ID "00001" is provided with a 300 yen coupon from A store. In order to prevent such biased calculation results from being output, the information processing device 100 uses the policy score in optimization calculations. Preventing biased calculation results from being output eliminates inequality in measures. Therefore, according to the embodiment, the information processing apparatus 100 can eliminate inequality in measures.
　また、上記では、施策として、クーポンを提供する場合を示した。例えば、施策として、ノベルティが提供されてもよい。 Additionally, the above example shows the case where coupons are provided as a measure. For example, novelty items may be provided as a measure.
　１００　情報処理装置、　１０１　プロセッサ、　１０２　揮発性記憶装置、　１０３　不揮発性記憶装置、　１０４　通信インタフェース、　１１０　記憶部、　１１１　行動履歴テーブル、　１１２　行動特徴テーブル、　１１３　属性テーブル、　１１４　施策結果テーブル、　１１５　学習済モデル、　１１６　施策候補テーブル、　１２０　学習部、　１３０　取得部、　１４０　生成部、　１５０　予測部、　１６０　決定部、　１７０　出力部、　２００　端末装置、　３００　学習データ。 100 Information processing device, 101 Processor, 102 Volatile storage device, 103 Non-volatile storage device, 104 Communication interface, 110 Storage unit, 111 Behavior history table, 112 Behavior feature table, 113 Attribute table, 11 4 Measure result table, 115 Learned Model, 116 Measure candidate table, 120 Learning unit, 130 Acquisition unit, 140 Generation unit, 150 Prediction unit, 160 Determination unit, 170 Output unit, 200 Terminal device, 300 Learning data.

Claims (3)

  1. 　学習済モデル、ユーザ毎の行動特徴を示す行動特徴情報、前記ユーザ毎の属性を示す属性情報、施策の候補を示す施策候補情報、施策の不平等を緩和するための値である施策スコア、及び予算を示す予算情報を取得する取得部と、
    　前記行動特徴情報、前記属性情報、及び前記施策候補情報に基づいて、データを生成する生成部と、
    　生成された前記データと前記学習済モデルとに基づいて、施策を行ったときの売上又は来店増加回数を予測する予測部と、
    　予測結果と前記施策スコアとを用いて、前記予算内で最適化計算を行い、前記ユーザ毎の施策を決定する決定部と、
    　を有する情報処理装置。     A trained model, behavioral characteristic information indicating behavioral characteristics of each user, attribute information indicating attributes of each user, policy candidate information indicating policy candidates, policy score which is a value for alleviating inequality of measures, and an acquisition unit that acquires budget information indicating the budget;
    a generation unit that generates data based on the behavior characteristic information, the attribute information, and the measure candidate information;
    a prediction unit that predicts an increase in sales or the number of store visits when the measure is implemented based on the generated data and the learned model;
    a determining unit that performs an optimization calculation within the budget using the prediction result and the measure score, and determines the measure for each user;
    An information processing device having:
  2. 　情報処理装置が、
    　学習済モデル、ユーザ毎の行動特徴を示す行動特徴情報、前記ユーザ毎の属性を示す属性情報、施策の候補を示す施策候補情報、施策の不平等を緩和するための値である施策スコア、及び予算を示す予算情報を取得し、前記行動特徴情報、前記属性情報、及び前記施策候補情報に基づいて、データを生成し、生成された前記データと前記学習済モデルとに基づいて、施策を行ったときの売上又は来店増加回数を予測し、
    　予測結果と前記施策スコアとを用いて、前記予算内で最適化計算を行い、前記ユーザ毎の施策を決定する、
    　決定方法。     The information processing device
    A trained model, behavioral characteristic information indicating behavioral characteristics of each user, attribute information indicating attributes of each user, policy candidate information indicating policy candidates, policy score which is a value for alleviating inequality of measures, and Acquire budget information indicating a budget, generate data based on the behavior characteristic information, the attribute information, and the measure candidate information, and implement measures based on the generated data and the learned model. Predict the increase in sales or number of store visits when
    performing optimization calculations within the budget using the prediction result and the measure score, and determining measures for each user;
    How to decide.
  3. 　情報処理装置に、
    　学習済モデル、ユーザ毎の行動特徴を示す行動特徴情報、前記ユーザ毎の属性を示す属性情報、施策の候補を示す施策候補情報、施策の不平等を緩和するための値である施策スコア、及び予算を示す予算情報を取得し、前記行動特徴情報、前記属性情報、及び前記施策候補情報に基づいて、データを生成し、生成された前記データと前記学習済モデルとに基づいて、施策を行ったときの売上又は来店増加回数を予測し、
    　予測結果と前記施策スコアとを用いて、前記予算内で最適化計算を行い、前記ユーザ毎の施策を決定する、
    　処理を実行させる決定プログラム。
          In the information processing device,
    A trained model, behavioral characteristic information indicating behavioral characteristics of each user, attribute information indicating attributes of each user, policy candidate information indicating policy candidates, policy score which is a value for alleviating inequality of measures, and Acquire budget information indicating a budget, generate data based on the behavior characteristic information, the attribute information, and the measure candidate information, and implement measures based on the generated data and the learned model. Predict the increase in sales or number of store visits when
    performing optimization calculations within the budget using the prediction result and the measure score, and determining measures for each user;
    A decision program that executes a process.
PCT/JP2022/031794 2022-08-24 2022-08-24 Information processing device, determination method, and determination program WO2024042627A1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/JP2022/031794 WO2024042627A1 (en) 2022-08-24 2022-08-24 Information processing device, determination method, and determination program

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/JP2022/031794 WO2024042627A1 (en) 2022-08-24 2022-08-24 Information processing device, determination method, and determination program

Publications (1)

Publication Number Publication Date
WO2024042627A1 true WO2024042627A1 (en) 2024-02-29

Family

ID=90012672

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2022/031794 WO2024042627A1 (en) 2022-08-24 2022-08-24 Information processing device, determination method, and determination program

Country Status (1)

Country Link
WO (1) WO2024042627A1 (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080249844A1 (en) * 2002-07-19 2008-10-09 International Business Machines Corporation System and method for sequential decision making for customer relationship management
JP2012003428A (en) * 2010-06-15 2012-01-05 Dainippon Printing Co Ltd Electronic coupon delivery device, electronic coupon delivery system and method
WO2015079460A1 (en) * 2013-11-28 2015-06-04 Gupta Lucky System for computing contribution and providing appropriate incentives
US20170300956A1 (en) * 2016-04-15 2017-10-19 Wal-Mart Stores, Inc. Systems and methods to generate coupon offerings to identified customers

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080249844A1 (en) * 2002-07-19 2008-10-09 International Business Machines Corporation System and method for sequential decision making for customer relationship management
JP2012003428A (en) * 2010-06-15 2012-01-05 Dainippon Printing Co Ltd Electronic coupon delivery device, electronic coupon delivery system and method
WO2015079460A1 (en) * 2013-11-28 2015-06-04 Gupta Lucky System for computing contribution and providing appropriate incentives
US20170300956A1 (en) * 2016-04-15 2017-10-19 Wal-Mart Stores, Inc. Systems and methods to generate coupon offerings to identified customers

Similar Documents

Publication Publication Date Title
AU2015317621B2 (en) Method and apparatus for predicting customer intentions
US11210300B2 (en) System and methods of generating structured data from unstructured data
US11694257B2 (en) Utilizing artificial intelligence to make a prediction about an entity based on user sentiment and transaction history
US10346752B2 (en) Correcting existing predictive model outputs with social media features over multiple time scales
US20210118004A1 (en) Method and system for personalizing offers
US20210182730A1 (en) Systems and methods for detecting non-causal dependencies in machine learning models
US20160125299A1 (en) Apparatus for data analysis and prediction and method thereof
Ahad Financial development and money demand function: Cointegration, causality and variance decomposition analysis for Pakistan
Hu et al. Tourism demand forecasting using nonadditive forecast combinations
Pinçe et al. The role of contract expirations in service parts management
KR20210144330A (en) Method and apparatus for recommending item based on collaborative filtering neural network
US20230032739A1 (en) Propensity modeling process for customer targeting
EP3374936A1 (en) Method and apparatus for dynamically selecting content for online visitors
US20170046726A1 (en) Information processing device, information processing method, and program
WO2024042627A1 (en) Information processing device, determination method, and determination program
US20220207324A1 (en) Machine-learning techniques for time-delay neural networks
US20220414693A1 (en) Automatic user retention system
JP2003323601A (en) Predicting device with reliability scale
Delgado et al. Simple Objectives Work Better.
US20230136972A1 (en) Egocentric network entity robustness prediction
Haubitz et al. How to change a running system—controlling the transition to optimized spare parts inventory policies
JP2021103339A (en) Device, method, and program for analyzing customer attribute information
JP7351887B2 (en) Information processing device, information processing system, and information processing method
US20230368130A1 (en) Systems and methods for prioritizing orders
JP7366218B1 (en) Information processing device, method and program

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22956455

Country of ref document: EP

Kind code of ref document: A1