Data processing device, inference device, machine learning device, data processing method, inference method, and machine learning method

Abstract

To provide a data processing apparatus and an inference apparatus that make it possible to accurately predict an interest behavior in a product of a person while taking into account a factor that motivates the person to take the behavior and an effect of measures implemented for the product.SOLUTION: A data processing apparatus 4 includes a data acquisition unit 400 that acquires input data including person data in which a behavior of a person of prediction target is recorded based on one or more standpoints, a behavior factor index indicating a factor that motivates the person to take the behavior, and measure data indicating measures of the prediction target performed for the person regarding a product, and a product interest prediction unit 401 that predicts a product interest behavior of the person of the prediction target regarding the products as a behavior caused by the measures, based on output data output from a learning model 12 by inputting input data acquired by the data acquisition unit 400 into the learning model 12 that has already been trained.SELECTED DRAWING: Figure 10

Description

[Technical Field]

[0001] The present invention relates to a data processing device, an inference device, a machine learning device, a data processing method, an inference method, and a machine learning method. [Background technology]

[0002] Various methods for predicting the sales status of a product have been developed, and the results of the prediction of the sales status of a product are utilized as useful information for, for example, product planning, product development, production management, purchasing and order management, sales strategy formulation, marketing, advertising, etc. As a system for predicting the sales status of a product, for example, Patent Document 1 discloses a system for predicting the number of sales of a product based on the past sales performance of the product and attention level information associated with the product. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent Publication No. 2021-103444 Summary of the Invention [Problem to be solved by the invention]

[0004] When each person takes product interest action toward a product (e.g., product purchase action, etc.), it is considered that each person has inherent behavioral factor indices (e.g., sensibilities, etc.) that indicate factors that motivate that action. Furthermore, it is considered that measures implemented by businesses regarding products affect each person's product interest action toward the product. However, the system disclosed in Patent Document 1 does not take into consideration such behavioral factor indices inherent in each person or measures implemented regarding the product, and does not predict product interest action toward a product based on behavioral factor indices or measures.

[0005] The present invention has been made in consideration of the above-mentioned problems, and aims to provide a data processing device, an inference device, a machine learning device, a data processing method, an inference method, and a machine learning method that enable accurate prediction of a person's product interest behavior while taking into account the factors that motivate a person to take action and the influence of measures implemented regarding the product. [Means for solving the problem]

[0006] In order to achieve the above object, a data processing device according to one aspect of the present invention comprises: a data acquisition unit that acquires input data including person data that records the behavior of a person to be predicted based on one or more viewpoints, behavioral factor indicators that indicate factors that motivate the person to take the behavior, and action data that indicates a prediction target action to be implemented for the person regarding a product; The input data acquired by the data acquisition unit includes the person data that records the behavior of the person to be learned based on the viewpoint, the behavioral factor index that indicates the factors that motivate the person to take the behavior, and the policy data that indicates the policy to be learned that will be implemented for the person in relation to the product, and the output data includes product interest behavior data that records product interest behavior toward the product as the behavior caused by the person as a result of the policy, and a product interest behavior prediction unit that predicts the product interest behavior of the person to be predicted based on the output data output from the learning model that has been trained by machine learning. [Effects of the Invention]

[0007] In a data processing device according to one aspect of the present invention, a product interest behavior prediction unit inputs input data, including person data recording the behavior of a person to be predicted based on one or more perspectives, behavioral factor indices indicating factors motivating the person to take action, and policy data indicating measures to be implemented for the person regarding a product, into a learning model, and predicts the product interest behavior of the person as a behavior caused by the measures, based on output data output from the learning model. Thus, the product interest behavior of a person can be accurately predicted while taking into account the influence of the factors motivating the person to take action and the measures implemented regarding the product.

[0008] Problems, configurations, and effects other than those described above will become apparent from the detailed description of the invention that follows. [Brief explanation of the drawings]

[0009] [Figure 1] 1 is an overall configuration diagram showing an example of a policy planning system 1. FIG. [Figure 2] FIG. 9 is a hardware configuration diagram showing an example of a computer 900. [Figure 3] FIG. 2 is a data configuration diagram showing an example of a person database 20. [Figure 4] FIG. 2 is a data configuration diagram showing an example of a policy database 21. [Figure 5] FIG. 2 is a data configuration diagram showing an example of an external environment database 22. [Figure 6] FIG. 2 is a block diagram showing an example of a machine learning device 3. [Figure 7] 1 is a diagram showing an example of learning data 11 and a learning model 12. FIG. [Figure 8] FIG. 2 is a diagram showing the relationship between each of the databases 20 to 22 and the learning data 11. [Figure 9] FIG. 2 is a block diagram showing an example of a data processing device 4. [Figure 10] FIG. 2 is a functional explanatory diagram showing an example of a data processing device 4. [Figure 11]10 is a flowchart showing an example of a machine learning method performed by the machine learning device 3. [Figure 12] 10 is a flowchart showing an example of a data processing method (policy evaluation process) performed by the data processing device 4. [Figure 13] 10 is a flowchart showing an example of a data processing method (measure extraction process) by the data processing device 4. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, an embodiment for carrying out the present invention will be described with reference to the drawings. The scope necessary for the explanation to achieve the object of the present invention will be schematically shown, and the scope necessary for explaining the relevant parts of the present invention will be mainly explained, and the parts that are omitted from the explanation will be based on publicly known techniques.

[0011] (Policy Planning System 1) Figure 1 is an overall configuration diagram showing an example of a policy planning system 1. The policy planning system 1 functions as a system for evaluating and planning policies by predicting a person's product interest behavior based on person data that records the person's behavior, behavioral factor indicators (tags) that indicate the factors that motivate the person to take action, policy data that indicates policies to be implemented for the person regarding products, and external environment data that records the state of the external environment and events that occur in the external environment.

[0012] Person data is data that records a person's behavior based on one or more perspectives. Perspectives for recording a person's behavior include, for example, purchasing behavior when a person purchases a product at a store or an e-commerce site, web browsing behavior when a person browses a website, web search behavior when a person searches on a website, travel behavior when a person travels by any means of transportation (walking, train, car, etc.), and survey response behavior when a person responds to a survey on a website or in print. Therefore, person data includes, for example, purchasing behavior, This includes purchase history that records web browsing behavior, web browsing history that records web search behavior, travel history that records travel behavior, and survey response history that records survey response behavior.

[0013] Behavioral factor indices indicate factors that motivate a person to take a specific action. There are multiple types of behavioral factor indices classified by, for example, sensibility, preference, attribute, etc. Examples of sensibility and preference include health-conscious, organic, preference for origin of produce, price (savings) conscious, luxury-conscious, safety and security, high sensitivity to beauty, cleanliness freak, preference for Japanese food, preference for Western food, sweet tooth (sweet tooth), spicy food (spicy tooth), strong flavor lover, sour flavor lover, carbonated food lover, indoor type, outdoor type, etc. Examples of attributes related to demographics and lifestyle include age, gender, single, married, with children, owning a home, renting, owning a car, keeping pets, being on a diet, etc.

[0014] A behavioral factor index is assigned to a person based on the value of a variable, and a different variable value is assigned for each type of behavioral factor index. In this embodiment, the case where the behavioral factor index is defined using a two-class classification (binary classification of 0 or 1) will be mainly described. For example, the variable values ​​of the behavioral factor index for "health-consciousness" are assigned as "0: not health-conscious" and "1: health-conscious." Note that the behavioral factor index may also be defined using a multi-class classification (multi-value classification). For example, when the behavioral factor index is defined using a three-class classification, the variable values ​​of the behavioral factor index for "health-consciousness" may be assigned as "-1: not health-conscious," "0: neutral," and "1: health-conscious." Furthermore, the definition of the behavioral factor index may differ for each type of behavioral factor index. Furthermore, instead of or in addition to the classification values ​​for each type of behavioral factor index (which may be either binary or multi-value classification), the behavioral factor index may be assigned as a feature value quantified by processing or aggregating specific behaviors that a person has performed in the past.

[0015] The campaign data indicates campaigns implemented for a specific person (or a group of people including multiple people) regarding a product. A product is a target of a person's product interest behavior, and includes not only goods but also services. Some or all of the products may be assigned product indicators that indicate the characteristics of the product and correspond to the behavioral factor indicators. For example, a product "cake" made with ingredients that are particular about the origin of the ingredients may be assigned product indicators such as "sweet tooth" or "particular about origin," while a product "cake" made with high-quality ingredients may be assigned product indicators such as "sweet tooth" or "luxury-oriented." Campaigns are implemented primarily for the purpose of promoting the sales of a product, and examples include sales, advertisements, and promotional events. Therefore, the campaign data may include, for example, the content of the campaign implemented for the product, and, if a product indicator is assigned to the product, the product indicator may also be included.

[0016] Product interest behavior is any behavior that a person takes when they are interested in a product. Product interest behavior includes, for example, purchasing behavior in which a person purchases a product at a store or on an e-commerce site, purchase reservation behavior in which a person reserves the purchase of a product, candidate registration behavior in which a person registers a product as a candidate (favorite) for purchase, information gathering behavior in which a person orders or views catalogs or service information materials, store visit behavior in which a person visits a store or on an e-commerce site, event participation behavior in which a person participates in a promotional event related to a product, and product recommendation behavior in which a person recommends a product to another person.

[0017] External environment data is data that records the state of the external environment and events that have occurred in the external environment based on one or more perspectives. Perspectives for recording the external environment include, for example, weather at each time (or period) or location (or region), trends, external events, etc. Therefore, external environment data may include, for example, meteorological data including weather, temperature, humidity, precipitation, snowfall, ultraviolet rays, pollen, etc., various rankings, SNS (social networking) data, etc. This includes trending data, including topics discussed on social media platforms (e.g., social networking services), and external event data, including details of external events from entertainment, sports, public service, etc.

[0018] The policy planning system 1 comprises, as its main components, a database device 2, a machine learning device 3, a data processing device 4, a worker terminal device 5, and a policy planner terminal device 6. Each of the devices 2 to 6 is configured, for example, as a general-purpose or dedicated computer (see FIG. 2 described below), and is connected to a wired or wireless network 7 so that various data can be transmitted and received between them. Note that the number of the devices 2 to 6 and the connection configuration of the network 7 are not limited to the example in FIG. 1 and may be changed as appropriate.

[0019] The database device 2 includes a person database 20 capable of registering person data and behavioral factor indexes for each person, a policy database 21 capable of registering policy data for each policy, and an external environment database 22 capable of registering external environment data.

[0020] The person database 20 cooperates with an external system (not shown) that records each person's daily actions as an action history based on one or more perspectives, thereby registering action histories based on one or more perspectives as person data for each person. The policy database 21 cooperates with an external system (not shown) that records actions implemented by businesses and the like as a policy history, thereby registering policy histories as policy data for each policy. The external environment database 22 cooperates with an external system (not shown) that records the external environment based on one or more perspectives, thereby registering external environment data based on one or more perspectives. The external system is a system that provides various information to the database device 2 in a push or pull manner, and is composed of, for example, a point-of-sale information management system (POS system), a computer operation log collection system, a navigation system, a questionnaire system, an advertisement distribution server, a weather distribution server, a trend distribution server, an event distribution server, etc.

[0021] The machine learning device 3 operates as the main subject of the learning phase of machine learning. The machine learning device 3 performs machine learning of a learning model 12 by combining person data and behavioral factor indicators registered in a person database 20, policy data registered in a policy database 21, and external environment data registered in an external environment database 22, and using these as learning data 11. The trained learning model 12 is provided to the data processing device 4 via a network 7, a recording medium, or the like. Supervised learning is adopted as the machine learning method.

[0022] The data processing device 4 operates as the main body of the inference phase of machine learning. Using the trained learning model 12 generated by the machine learning device 3, the data processing device 4 predicts a person's product interest behavior, evaluates and extracts measures, and provides the results to the database device 2, the measure planner terminal device 6, etc.

[0023] The worker terminal device 5 is a terminal device used by a worker 10A who generates the learning data 11 and the learning model 12 when the machine learning device 3 performs machine learning of the learning model 12. The worker terminal device 5 accepts various input operations via a display screen such as an application program or a web browser, and displays various information via the display screen.

[0024] The policy planner terminal device 6 is a terminal device used by the policy planner 10B who plans policies when the data processing device 4 predicts a person's product interest behavior and evaluates and extracts policies. The policy planner terminal device 6 accepts various input operations via a display screen such as an application program or a web browser, and displays various information via the display screen.

[0025] (Computer 900) 2 is a hardware configuration diagram showing an example of the computer 900. Each of the devices 2 to 6 of the policy planning system 1 is configured by a general-purpose or dedicated computer 900.

[0026] 2, the computer 900 includes, as its main components, a bus 910, a processor 912, a memory 914, an input device 916, an output device 917, a display device 918, a storage device 920, a communication I / F (interface) unit 922, an external device I / F unit 924, an I / O (input / output) device I / F unit 926, and a media input / output unit 928. Note that the above components may be omitted as appropriate depending on the application of the computer 900.

[0027] The processor 912 is composed of one or more arithmetic processing devices (such as a central processing unit (CPU), a micro-processing unit (MPU), a digital signal processor (DSP), or a graphics processing unit (GPU)), and operates as a control unit that controls the entire computer 900. The memory 914 stores various data and programs 930, and is composed of, for example, a volatile memory (such as a DRAM or SRAM) that functions as a main memory, a non-volatile memory (ROM), a flash memory, etc.

[0028] The input device 916 is composed of, for example, a keyboard, a mouse, a numeric keypad, an electronic pen, etc., and functions as an input unit. The output device 917 is composed of, for example, a sound (audio) output device, a vibration device, etc., and functions as an output unit. The display device 918 is composed of, for example, a liquid crystal display, an organic EL display, electronic paper, a projector, etc., and functions as an output unit. The input device 916 and the display device 918 may be integrated into one device, such as a touch panel display. The storage device 920 is composed of, for example, a hard disk drive (HDD), a solid state drive (SSD), etc., and functions as a storage unit. The storage device 920 stores various data necessary for executing the operating system and the program 930.

[0029] The communication I / F unit 922 is connected to a network 940 such as the Internet or an intranet (which may be the same as network 7 in FIG. 1) via a wired or wireless connection and functions as a communication unit that transmits and receives data to and from other computers in accordance with a predetermined communication protocol. The external device I / F unit 924 is connected to an external device 950 such as a camera, printer, scanner, or reader / writer via a wired or wireless connection and functions as a communication unit that transmits and receives data to and from the external device 950 in accordance with a predetermined communication protocol. The I / O device I / F unit 926 is connected to an I / O device 960 such as various sensors and actuators and functions as a communication unit that transmits and receives various signals and data, such as detection signals from sensors and control signals to actuators, to and from the I / O device 960. The media input / output unit 928 is formed by a drive device such as a DVD (Digital Versatile Disc) drive or a CD (Compact Disc) drive and reads and writes data from and to media (non-transitory storage media) 970 such as DVDs and CDs.

[0030] In the computer 900 having the above configuration, the processor 912 loads the program 930 stored in the storage device 920 into the memory 914, executes it, and controls each unit of the computer 900 via the bus 910. The program 930 may be stored in the memory 914 instead of the storage device 920. The program 930 may be recorded on the medium 970 in an installable file format or an executable file format, and provided to the computer 900 via the media input / output unit 928. The program 930 may be provided to the computer 900 by downloading it via the communication I / F unit 922 over the network 940. The computer 900 also includes a processor 912. The various functions realized by the CPU 12 executing the program 930 may be realized by hardware such as a field-programmable gate array (FPGA) or an application specific integrated circuit (ASIC).

[0031] The computer 900 is an electronic device of any type, such as a desktop computer or a portable computer, and may be a client computer, a server computer, or a cloud computer.

[0032] (People Database 20) 3 is a data configuration diagram showing an example of the person database 20. The person database 20 is a database for managing person data and behavioral factor indicators based on person identification information (user ID) assigned to each person. The person database 20 is composed of, for example, a purchase history table, a web browsing history table, a web search history table, a movement history table, a questionnaire response history table, a behavioral factor indicator table, product master information, store master information, website master information, and map information.

[0033] The purchase history table has multiple records for recording purchasing behavior and purchase reservation behavior (an example of product interest behavior), and each record registers date and time, product, price, store (which may be an e-commerce site), etc. A product ID is registered for a product, and the product is linked to product master information by the product ID, thereby associating the product with product attributes (product name, product category, price (suggested retail price, open price, etc.)) registered in the product master information. A store ID is registered for a store, and the store is linked to store master information by the store product ID, thereby associating the store with store attributes (store name, store category, etc.) registered in the store master information. Note that a product may be assigned a product index (behavioral characteristic index) corresponding to the behavioral factor index, and may be registered in the product master information as part of the product attribute, for example. A store may be assigned a store index (behavioral characteristic index) corresponding to the behavioral factor index, and may be registered in the store master information as part of the store attribute, for example.

[0034] The web browsing history table has multiple records for recording web browsing behavior, including, for example, candidate registration behavior, information gathering behavior, store visit behavior, and product recommendation behavior (an example of product interest behavior), with each record registering the date and time, website, browsing time, etc. The web search history table has multiple records for recording web search behavior, with each record registering the date and time, search words, etc. A URL for accessing a website is registered, and the URL is linked to website master information, thereby associating the website with website attributes (administrator name, website category, etc.) registered in the website master information. Note that a website may be assigned a website index (behavioral feature index) corresponding to a behavioral factor index, and may be registered in the website master information as part of the website attributes, for example.

[0035] The movement history table has a plurality of records for recording movement behaviors including, for example, store visit behaviors and event participation behaviors (an example of product interest behaviors), and each record registers date and time, location, means of transportation, etc. For example, a stay point or area where a user stayed for a predetermined period of time or more is registered as a location, and the stay point or area is linked to map information, thereby associating the location attributes (location name, location category, etc.) and area attributes (area name, area category, etc.) registered in the map information. Note that the locations and areas may be assigned location indices (behavioral characteristic indices) or area indices (behavioral characteristic indices) corresponding to the behavioral factor indices, and may be registered in the map information as part of the location attributes or area attributes, for example.

[0036] The survey response history table contains multiple records for recording survey response behavior. Each record contains the date and time, the question, the answer, etc.

[0037] The behavioral factor indicator table has multiple records for recording behavioral factor indicators assigned to a person, and each record registers the date and time, the value of the variable of the behavioral factor indicator (0 or 1 in the case of binary classification), etc. In the example of FIG. 3, three fields are shown corresponding to three behavioral factor indicators, "health-conscious," "sweet tooth," and "car ownership," but similar fields are also provided for other behavioral factor indicators. Furthermore, as the date and time, the date and time when each behavioral factor indicator was changed is registered; for example, in the case of binary classification, the date and time when "1" was assigned and the date and time when "0" was assigned are registered.

[0038] In the person database 20, the information in each table is associated with a user ID, so that person data recording the person's behavior based on multiple perspectives and behavioral factor indices are managed for each person. The information registered in the person database 20 is configured to enable various statistical processes to be performed. Note that the person data may be recorded in association with feature data (in this embodiment, products, stores, websites, locations, and regions) to which behavioral feature indices (in this embodiment, product indices, store indices, website indices, location indices, and region indices) indicating the characteristics of behavior and corresponding to the behavioral factor indices are assigned. In addition to the above, behavioral feature indices may also be assigned to characteristics such as time of day, day of the week, and season.

[0039] The behavioral factor index for a person and the behavioral feature index for a product, store, website, location, and region may be assigned by, for example, displaying the person data registered in the person database 20 on the worker terminal device 5, and the worker 10A confirming the displayed person data and performing an input operation (annotation work), or by having the data processing device 4 reference the person data registered in the person database 20 and perform predetermined data processing on the person data. The predetermined data processing may use a decision rule or a learning model, for example, and may be based on the technology disclosed in Japanese Patent Application No. 2021-213643 filed by the applicant of the present application. The entire contents of the specification and drawings of Japanese Patent Application No. 2021-213643 are incorporated herein by reference. The behavioral factor index for a person and the behavioral feature index for a product, store, website, location, and region may be assigned as a quantified feature by processing the person data registered in the person database 20 according to predetermined processing conditions or aggregating according to predetermined aggregation conditions.

[0040] (Policy Database 21) 4 is a data structure diagram showing an example of the policy database 21. The policy database 21 is a database for managing policy data based on policy identification information (policy ID) assigned to each policy. The policy database 21 is composed of, for example, a sales table, an advertisement table, and a sales promotion event table.

[0041] The sale table has multiple records for recording information about product sales, and each record registers the date and time, product, store, sale details, target users for the sale, etc. A product ID is registered for each product, and the product is linked to product master information by the product ID, thereby associating the product with product attributes (product name, product category, product index, etc.) registered in the product master information. A store ID is registered for each store, and the store is linked to store master information by the store product ID, thereby associating the store with store attributes (store name, store category, store index, etc.) registered in the store master information. Sale details include, for example, the product's selling price, discount rate, discount amount, sale period, etc. A person ID and characteristics (which may be behavioral factor indexes) indicating the person who is the target of the sale are registered for the target users for the sale.

[0042] The advertisement table has a plurality of records for recording information about product advertisements, and each record contains information such as the date and time, product, store, advertising medium, advertisement content, and target users. Advertisements are carried out through any advertising medium, such as television, radio, newspapers, flyers, magazines, public transport, direct mail, listings, banners, video sites, and social media. The type of advertising medium, advertising frequency, advertising period, etc. are registered as advertising content. The person ID and characteristics (which may be behavioral factor indicators) that indicate the person to whom the advertisement is directed are registered as advertising target users. As with the sales table, product IDs and store IDs are registered for products and stores, respectively.

[0043] The promotional event table has multiple records for recording information about promotional events such as demonstration sales, exhibitions, gift campaigns, fan services, etc., and each record registers the date and time, product, store, event content, target user for the event, etc. The event content registers the type of promotional event, gift details (product, price), etc. The event target user registers a person ID and characteristics (which may be behavioral factor indicators) indicating the person targeted by the promotional event. As with the sales table, the product ID and store ID are registered for the product and store, respectively.

[0044] (External Environment Database 22) 5 is a data configuration diagram showing an example of the external environment database 22. The external environment database 22 is a database for managing external environment data, and is made up of, for example, a weather table, a trend table, and an external event table.

[0045] The weather table has, for example, a plurality of records for recording information about the weather, and each record registers the date and time, location, weather, temperature, humidity, amount of precipitation, amount of snowfall, ultraviolet rays, pollen, etc.

[0046] The trend table has multiple records for recording information about trends such as various rankings and topics on social media, and each record registers the date and time, location, trend content, event target user, etc. Trend content registers the type and tendency of the trend, etc. Trend target user registers a person ID and characteristics (which may be behavioral factor indicators) that indicate the person who is the target of the trend.

[0047] The external event table has multiple records for recording information about external events such as entertainment, sports, and public services, and each record registers the date and time, location, event content, event target user, etc. The event content registers the type of external event, etc. The event target user registers a person ID and characteristics (which may be behavioral factor indicators) indicating the person who is the target of the external event.

[0048] The data structure of each of the databases 20 to 22 may be changed as appropriate, and data other than those described above may be registered, or some of the data described above may be omitted. Also, while Figures 3 to 5 show the fields possessed by each of the databases 20 to 22, depending on the record, it is not necessary to register all field values, and some field values ​​may be omitted.

[0049] (Machine Learning Device 3) 6 is a block diagram showing an example of a machine learning device 3. The machine learning device 3 includes a control unit 30 configured with a processor or the like, a storage unit 31 configured with an HDD, an SSD, a memory or the like, a communication unit 32 that is a communication interface with the network 7, an input unit 33 configured with a keyboard, a mouse or the like, and a display unit 34 configured with a display or the like. Note that the input unit 33 and the display unit 34 may be omitted.

[0050] The memory unit 31 stores the learning data 11, the learning model 12, and the machine learning program 310. The storage unit 31 stores the learning data 11, as well as the operating system, other programs, various data, etc. The storage unit 31 functions as a learning data storage unit that stores the learning data 11, and a trained model storage unit that stores the training model 12.

[0051] As shown in FIG. 6, the control unit 30 executes a machine learning program 310 stored in the storage unit 41, thereby functioning as a learning data acquisition unit 300 and a machine learning unit 301.

[0052] When the learning data acquisition unit 300 receives acquisition conditions for the learning data 11 from the worker terminal device 5 as an input operation by the worker 10A, the learning data acquisition unit 300 acquires the learning data 11 consisting of input data and output data by referring to each of the databases 20 to 22 based on the acquisition conditions. Then, the learning data acquisition unit 300 provides the multiple sets of learning data 11 to the machine learning unit 301 or stores them in the storage unit 31.

[0053] The machine learning unit 301 inputs multiple sets of learning data 11, each consisting of input data and output data, into the learning model 12, causing the learning model 12 to learn the correlation between the input data and the output data through machine learning. Then, the machine learning unit 301 stores the learned learning model 12 (specifically, a set of adjusted parameters) in the storage unit 31. Note that when performing machine learning, the machine learning unit 301 can employ any method, such as online learning, batch learning, or mini-batch learning.

[0054] 7 is a diagram showing an example of the learning data 11 and the learning model 12. FIG. 8 is a diagram showing the relationship between the learning data 11 and each of the databases 20 to 22.

[0055] The learning data 11 is composed of input data and output data. The learning data 11 is data used as teacher data (training data), verification data, and test data in supervised learning. The output data included in the learning data 11 is data used as a correct answer label in supervised learning.

[0056] As shown in Figure 7, the input data includes person data that records the behavior of the person being studied based on one or more perspectives, behavioral factor indicators that indicate the factors that motivate the person to take action, policy data that indicates the measures to be implemented for the person being studied regarding the product, and external environment data that indicates the external environment at the time the measures are implemented.

[0057] The output data includes product interest behavior data that records product interest behavior regarding a product as a behavior caused by the person being studied as a result of the measure being studied, as shown in Fig. 7. The product interest behavior data includes, for example, at least one of purchasing behavior indicating whether or not the person purchased a product, purchase reservation behavior indicating whether or not the person reserved the purchase of a product, candidate registration behavior indicating whether or not the person registered a product as a candidate for purchase, information gathering behavior indicating whether or not the person ordered or viewed a catalog or service information material, store visit behavior indicating whether or not the person visited a store or an e-commerce site, event participation behavior indicating whether or not the person participated in a sales promotion event, and product recommendation behavior indicating whether or not the person recommended a product to another person.

[0058] The person data included in the input data is person data that records the behavior of the person being studied at the learning reference point or a period prior to the learning reference point, when the time when the policy being studied is implemented is defined as the learning reference point (see Figure 8), and is obtained from the person database 20.

[0059] The behavioral factor indexes included in the input data are behavioral factor indexes assigned to the person to be learned at the learning reference point in time, and are acquired from the person database 20. In the example of FIG. At the Xi baseline point in time, the behavioral factor indicators "sweet tooth" and "car ownership" have been assigned, so "1" is entered for the behavioral factor indicators "sweet tooth" and "car ownership," and "0" is entered for "health-conscious."

[0060] The policy data included in the input data is policy data indicating the policies implemented at the learning reference point in time, and is acquired from the policy database 21. In the example of Fig. 8, the policy data records that the policy to be learned is "sale," and that the sale content is "20% discount rate" for the product "cake" to which the product indicators of "sweet tooth" and "preference for origin" are assigned, for example.

[0061] The external environment data included in the input data indicates the external environment at the learning reference time point, and is acquired from the external environment database 22. In the example of Fig. 8, the external environment data includes weather data at the learning reference time point and at the location of the store included in the policy data (learning reference point).

[0062] The product interest behavior data included in the output data is product interest behavior data that records the product interest behavior (e.g., purchasing behavior, etc.) of the person to be studied at the learning reference point in time or a period in the future than the learning reference point in time, and is acquired from each table (e.g., purchase history table, etc.) of the person database 20. As described above, the product interest behavior data includes at least one of purchasing behavior, purchase reservation behavior, candidate registration behavior, information gathering behavior, store visit behavior, event participation behavior, product recommendation behavior, etc., and for example, if a product interest behavior has occurred, "1" is recorded, and if a product interest behavior has not occurred, "0" is recorded. The examples of FIGS. 7 and 8 show a case where the product interest behavior data records purchasing behavior ("1") in which the person to be studied has purchased a product.

[0063] When the learning data acquiring unit 300 receives, as an acquisition condition for the learning data 11, for example, a previously implemented policy to be learned (in the example of FIG. 8 , a “sale” policy), the learning data acquiring unit 300 acquires policy data (part of the input data) indicating the policy to be learned by referring to the policy database 21, sets the product in the policy data as the product to be learned, sets the target user in the policy data as the person to be learned, sets the date and time of the policy in the policy data as the learning reference time point, and sets the location of the store in the policy data as the learning reference point (setting of the learning reference point may be omitted). Then, the learning data acquiring unit 300 acquires person data (part of the input data) of the person to be learned for a period prior to the learning reference time point by referring to the person database 20, and acquires behavioral factor indicators (part of the input data) of the person to be learned at the learning reference time point. Furthermore, the data acquiring unit 400 acquires external environment data (part of the input data) at the learning reference time point and the learning reference point by referring to the external environment database 22. Furthermore, the learning data acquisition unit 300 refers to each table in the person database 20 to acquire, as product interest behavior data (output data), information indicating whether or not the person being studied has exhibited product interest behavior with respect to the product being studied in a period future than the learning reference point (in the example of purchasing behavior shown in Figures 7 and 8, "1" if purchased, and "0" if not purchased).

[0064] The learning data acquisition unit 300 generates learning data 11 as shown by the dashed line in Fig. 8 by combining the input data and output data acquired by the above series of processes. In this case, if there are multiple target users, the learning data acquisition unit 300 performs the above series of processes for each person to generate multiple sets of learning data 11. Furthermore, if multiple measures to be learned are accepted as acquisition conditions for the learning data 11, the learning data acquisition unit 300 performs the above series of processes for each measure to generate multiple sets of learning data 11.

[0065] The learning model 12 employs, for example, a neural network structure, and includes an input layer 120, an intermediate layer 121, and an output layer 122, as shown in Fig. 7. Synapses (not shown) that connect the neurons are laid between the layers, and each synapse is associated with a weight.

[0066] The input layer 120 has neurons in a number corresponding to the input data of person data, behavioral factor indicators, policy data, and external environment data (each data may be a feature converted from each data in preprocessing), and each value of the input data is input to each neuron. The output layer 122 has neurons corresponding to product interest behavior data as output data, and prediction results (inference results) of product interest behavior for the input data are output as output data (values ​​ranging from 0 to 1 in this embodiment). Machine learning of the learning model 12 is performed by comparing the output data (correct answer labels) constituting the learning data 11 with the output data (inference results) output from the output layer 122 and performing a process (backpropagation, etc.) to adjust a group of parameters such as the weights of each synapse.

[0067] When inputting the input data constituting the training data 11 into the input layer 120 of the training model 12, the machine learning unit 301 may perform preprocessing, such as one-hot encoding or label encoding, to convert the input data into predetermined features to be input to the training model 12. In this case, the preprocessing may involve, for example, converting the purchase history, web browsing history, web search history, travel history, and survey response history included in the person data of the input data into features by appropriately combining various numerical operations with comparisons, ratios, frequencies, periods, categories, behavioral characteristic indices (product indices, store indices, website indices, location indices, and region indices). Similarly to the person data, the policy data and external environment data of the input data may also be converted into features by appropriately combining various numerical operations.

[0068] For example, if the person data includes feature data indicating behavioral characteristics (products, stores, websites, locations, and regions) and behavioral feature indices (product index, store index, website index, location index, and region index) associated with the feature data, as shown in Fig. 3, the machine learning unit 301 may perform preprocessing to convert the person data into features based on the behavioral feature indices assigned to the feature data included in the person data. In this case, the preprocessing may, for example, aggregate the purchase amount of products assigned with a behavioral factor index of "health-conscious," the purchase amount of products assigned with a behavioral factor index of "sweet tooth," the purchase amount of products assigned with a behavioral factor index of "car ownership," etc., based on product indices assigned to products previously purchased by the user that are recorded as part of the purchase history, and then normalize each of the aggregated purchase amounts to a value ranging from 0 to 1 to convert them into features.

[0069] 4, if the campaign data includes details of a campaign implemented for a product (date and time, product, store, sale details, advertisement details, event details, etc.) and product indices associated with the product, the machine learning unit 301 may perform preprocessing to convert the campaign data into features based on the details of the campaign included in the campaign data and the product indices assigned to the product. In this case, as preprocessing, for example, if the campaign details are "20% discount rate" and the product indices "sweet tooth" and "pick on origin" are assigned to the product, the product indices may be converted into features by setting the product indices "sweet tooth" and "pick on origin" to "0.2" and the other product indices to "0", respectively.

[0070] The number and types of learning models 12 that are machine-learned by the machine learning unit 301 and stored in the storage unit 31 may be changed as appropriate. For example, the machine learning method, A plurality of learning models 12 may be stored that have different conditions, such as the type of data, the type of data included in the behavioral factor index, the type of data included in the policy data, the type of data included in the external environment data, the method of preprocessing the input data, the type of data included in the product interest behavior data, etc. In this case, a plurality of types of learning data 11 having data configurations respectively corresponding to the plurality of learning models 12 with different conditions may be used.

[0071] (Data processing device 4) Fig. 9 is a block diagram showing an example of the data processing device 4. Fig. 10 is a functional explanatory diagram showing an example of the data processing device 4. The data processing device 4 includes a control unit 40 constituted by a processor or the like, a storage unit 41 constituted by an HDD, an SSD, a memory or the like, a communication unit 42 which is a communication interface with the network 7, an input unit 43 constituted by a keyboard, a mouse or the like, and a display unit 44 constituted by a display or the like. Note that the input unit 43 and the display unit 44 may be omitted.

[0072] The memory unit 41 stores the learning model 12 and the data processing program 410, as well as an operating system, other programs, various data, etc. The memory unit 41 functions as a trained model memory unit that stores the trained learning model 12.

[0073] The control unit 40 executes a data processing program 410 stored in the memory unit 41 to function as a data acquisition unit 400, a product interest behavior prediction unit 401, a policy evaluation unit 402, a policy extraction unit 403, and an output processing unit 404.

[0074] For example, when the data acquisition unit 400 receives prediction conditions for product interest behavior including the person to be predicted and the measure to be predicted from the worker terminal device 5 as an input operation by the measure planner 10B, or when it receives the prediction conditions from the measure evaluation unit 402 and the measure extraction unit 403 by working in cooperation with the measure evaluation unit 402 and the measure extraction unit 403, the data acquisition unit 400 acquires input data including person data, behavioral factor indexes, measure data, and external environment data based on the prediction conditions.

[0075] Specifically, by specifying the details of the campaign (date and time, product, store, sale details, advertisement details, event details, etc.) as the campaign to be predicted, the data acquisition unit 400 acquires campaign data indicating the campaign to be predicted based on the details of the campaign. The data acquisition unit 400 also sets the product in the campaign to be predicted as the product to be predicted, sets the date and time of the campaign to be predicted as the prediction reference time point, and sets the location of the store in the campaign to be predicted as the prediction reference position (setting of the prediction reference position may be omitted). The data acquisition unit 400 then refers to the person database 20 to acquire person data (part of the input data) of the person to be predicted for a period prior to the prediction reference time point. Furthermore, the data acquisition unit 400 refers to the external environment database 22 to acquire external environment data (part of the input data) at the prediction reference time point and the prediction reference position.

[0076] Through the above series of processes, the data acquisition unit 400 acquires input data including person data, behavioral factor indexes, policy data, and external environment data for the prediction conditions. Then, the data acquisition unit 400 provides the input data of the prediction target to the product interest behavior prediction unit 401 or stores it in the storage unit 41.

[0077] The product interest behavior prediction unit 401 predicts a person's product interest behavior based on output data output from the learning model 12 by inputting input data including person data, behavior factor indexes, policy data, and external environment data (each piece of data may be a feature converted from each piece of data in preprocessing) acquired by the data acquisition unit 400 into the trained learning model 12 stored in the storage unit 41. Note that the example of FIG. 9 shows a case where, as product interest behavior, whether or not to take purchasing behavior, i.e., whether or not to purchase a product, is predicted in accordance with FIG. 7. There are.

[0078] In this case, the product interest behavior prediction unit 401 performs post-processing on the predicted result (inference result) of product interest behavior output as output data (in this embodiment, a value in the range of 0 to 1) from the learning model 12 by comparing the variable value of the output data (in FIG. 10, the value for purchasing behavior is "0.9") with a predetermined threshold (e.g., "0.5"). If the variable value is equal to or greater than the threshold, the prediction result of purchasing behavior (an example of product interest behavior) of "1: purchase" is output. If the variable value is less than the threshold, the prediction result of purchasing behavior (an example of product interest behavior) of "0: do not purchase" is output. Note that when inputting input data to the input layer 120 of the learning model 12, the product interest behavior prediction unit 401 may perform pre-processing similar to that of the machine learning unit 301, thereby inputting the pre-processed input data, i.e., the feature quantities converted from the input data in the pre-processing, to the learning model 12.

[0079] The policy evaluation unit 402 accepts a person group including multiple people to be predicted and a policy to be predicted as policy evaluation conditions, and in cooperation with the data acquisition unit 400 and the product interest behavior prediction unit 401, predicts the product interest behavior of each person included in the person group and evaluates the effectiveness of the policy by statistically processing the product interest behavior of each person. The effectiveness of the policy is evaluated, for example, based on the product purchase rate, sales volume, sales amount, purchase reservation rate, purchase reservation amount, candidate registration rate, number of candidate registrations, information collection rate, information collection amount, store visit rate, number of store visitors, event participation rate, number of event participants, product recommendation rate, number of product recommenders, etc. The evaluation results of the policy effectiveness can be used in various services (product planning, product development, production management, purchasing / order management, sales strategy formulation, marketing, advertising, etc.).

[0080] The policy extraction unit 403 receives, as policy extraction conditions, a person group including multiple people to be predicted and a policy group including multiple policies to be predicted, and cooperates with the data acquisition unit 400 and the product interest behavior prediction unit 401 to predict the product interest behavior of each of the multiple people included in the person group for each of the multiple policies included in the policy group, and extracts policies that satisfy a predetermined criterion from the multiple policies by statistically processing the product interest behavior of each person for each policy. The predetermined criterion may be, for example, extracting a policy with the highest policy effect (product purchase rate, sales volume, sales amount, purchase reservation rate, purchase reservation amount, candidate registration rate, number of candidate registrations, information collection rate, information collection amount, store visit rate, number of store visitors, event participation rate, number of event participants, product recommendation rate, number of product recommenders, etc.), extracting a predetermined number of extractions in descending order of policy effect, or extracting policies whose effect exceeds a predetermined extraction criterion value.

[0081] The output processing unit 404 performs output processing for outputting the prediction result of product interest behavior by the product interest behavior prediction unit 401, the evaluation result of the measures by the measure evaluation unit 402, the extraction result of measures by the measure extraction unit 403, etc. to any output destination. Examples of output destinations include, but are not limited to, the database device 2, the measure planner terminal device 6, etc. In this case, the output processing unit 404 may output the prediction result of product interest behavior before post-processing (a value in the range of 0 to 1) as the prediction result by the product interest behavior prediction unit 401, in addition to or instead of the prediction result of product interest behavior after post-processing (0 or 1). Furthermore, the output processing unit 404 may process the evaluation result of measures and the extraction result of measures into the format of a table, a diagram, etc. and output the processed result.

[0082] The number of learning models 12 stored in the storage unit 41 is not limited to one, and multiple learning models 12 with different conditions, such as machine learning techniques, types of data included in person data, types of data included in behavioral factor indicators, types of data included in external environment data, methods of preprocessing input data, types of data included in product interest behavior data, etc. may be stored and selectively available. Also, the learning model 12 may be stored in the storage unit of an external computer (for example, a server-type computer or a cloud-type computer), and in that case, the data processing device 4 only needs to access the external computer. .

[0083] (machine learning methods) 11 is a flowchart showing an example of a machine learning method by the machine learning device 3. The following describes a case where an operator 10A performs an input operation using the operator terminal device 5, causing the machine learning device 3 to refer to each of the databases 20 to 22 and perform machine learning of the learning model 12.

[0084] First, in step S100, the worker terminal device 5 receives a machine learning instruction from the worker 10A, which includes an acquisition condition for the learning data 11. In step S101, the worker terminal device 5 transmits the machine learning instruction to the machine learning device 3. Note that the machine learning instruction may specify, for example, machine learning parameters and termination conditions.

[0085] Next, in step S110, when the machine learning device 3 receives the machine learning instruction for the learning model 12 sent in step S101, the learning data acquisition unit 300 refers to the databases 20 to 22 based on the acquisition conditions for the learning data 11 included in the machine learning instruction, and acquires multiple sets of learning data 11 (see Figures 7 to 8) that satisfy the acquisition conditions for the learning data 11.

[0086] Next, in step S120, the machine learning unit 301 uses the multiple sets of learning data 11 (which may be pre-processed learning data 11) acquired in step S110 to train the learning model 12 to learn the correlation between the input data (person data, behavioral factor index, policy data, and external environment data) and the output data (product interest behavior data).

[0087] Then, in step S130, the machine learning unit 301 stores the result in the storage unit 31 as a trained learning model 12. This completes the series of processes shown in FIG. 11. In the machine learning method, step S110 corresponds to a data acquisition step, step S120 corresponds to a machine learning step, and step S130 corresponds to a trained model storage step. Note that, although the above description assumes that the series of processes shown in FIG. 11 are executed when the worker terminal device 5 receives a machine learning instruction from the worker 10A, the series of processes shown in FIG. 11 may also be executed when the machine learning device 3 determines that a predetermined execution condition (for example, when a certain time has elapsed since the previous execution or when the person database 20 is updated) is satisfied.

[0088] As described above, the machine learning device 3 and machine learning method according to this embodiment can provide a learning model 12 that can predict (infer) a person's product interest behavior based on input data including person data, behavioral factor indicators, policy data, and external environment data. Therefore, by using this learning model 12 in the data processing device 4, it is possible to appropriately predict (infer) a person's product interest behavior.

[0089] (Data processing method) 12 and 13 are flowcharts showing an example of a data processing method by the data processing device 4. The following describes a case where a policy planner 10B performs an input operation using the policy planner terminal device 6, and the data processing device 4 evaluates policies (FIG. 12) and extracts policies (FIG. 13) based on the trained learning model 12.

[0090] 12, first, in step S200, the policy planner terminal device 6 receives a data processing instruction consisting of a person group including multiple people to be predicted and a policy to be predicted from the policy planner 10B, and then in step S201 transmits the data processing instruction to the data processing device 4. The person group may be, for example, a list of person IDs or conditions for creating a list of person IDs.

[0091] Next, in step S210, when the data processing device 4 receives the data processing instruction sent in step S201, the policy evaluation unit 402 accepts, based on the data processing instruction, a group of people including multiple people to be predicted and the policy to be predicted as policy evaluation conditions.

[0092] Then, in step S220, the policy evaluation unit 402 sequentially selects persons to be predicted from the group of persons, and performs a loop process of repeating steps S230 to S232 in cooperation with the data acquisition unit 400 and the product interest behavior prediction unit 401, thereby predicting the product interest behavior of each of the selected persons to be predicted.

[0093] In step S230, the data acquiring unit 400 receives from the policy evaluation unit 402 the prediction conditions for product interest behavior, including the person to be predicted selected in step S220 and the policy to be predicted received in step S210.

[0094] Then, in step S231, the data acquisition unit 400 refers to the person database 20 and the external environment database 22 based on the prediction conditions (the person to be predicted and the measure to be predicted), and acquires input data including person data, behavioral factor indicators, measure data, and external environment data for the prediction conditions.

[0095] Next, in step S232, the product interest behavior prediction unit 401 inputs the input data acquired in step S231 into the learning model 12, and predicts the person's product interest behavior based on the output data output from the learning model 12 as the inference result.

[0096] Next, in step S240, when the policy evaluation unit 402 receives the prediction results of product interest behavior for each person to be predicted from the product interest behavior prediction unit 401 as a result of the loop processing in step S220, the policy evaluation unit 402 evaluates the effectiveness of the policy by statistically processing the product interest behavior for each person.

[0097] Next, in step S250, the output processing unit 404 transmits screen information for displaying the effects of the measures on the display screen as the evaluation results of step S240 to the policy planner terminal device 6. At this time, the output processing unit 404 may include not only the evaluation results of the measures but also the prediction results of product interest behavior for each person in the screen information.

[0098] Next, in step S260, upon receiving the screen information, the policy planner terminal device 6 displays the evaluation results of the policy to be predicted on the display screen based on the screen information. This completes the series of processes shown in Fig. 12. In the data processing method, step S231 corresponds to a data acquisition step, step S232 corresponds to a product interest behavior prediction step, steps S210 to S240 correspond to a policy evaluation step, and step S250 corresponds to an output processing step.

[0099] 13, first, in step S300, the policy planner terminal device 6 receives a data processing instruction from the policy planner 10B, which includes a person group including multiple people to be predicted and a policy group including multiple policies to be predicted. In step S301, the policy planner terminal device 6 transmits the data processing instruction to the data processing device 4. The person group may be, for example, a list of person IDs, or conditions for creating a list of person IDs. The policy group may be a list of multiple policies with different policy contents, or conditions for creating a list of policies.

[0100] Next, in step S310, when the data processing device 4 receives the data processing instruction transmitted in step S301, the policy extraction unit 403 extracts a person group including a plurality of people to be predicted and a policy group including a plurality of policies to be predicted based on the data processing instruction, and calculates a policy evaluation condition. Accept as.

[0101] Next, in step S320, the policy extraction unit 403 sequentially selects policies to be predicted from the group of policies, and performs a loop process of repeating step S321. Then, in step S321, the policy extraction unit 403 sequentially selects people to be predicted from the group of people, and performs a loop process of repeating steps S330 to S332 in cooperation with the data acquisition unit 400 and the product interest behavior prediction unit 401, thereby predicting the product interest behavior of each of the selected people to be predicted.

[0102] In step S330, the data acquiring unit 400 receives, from the measure extracting unit 403, a prediction condition for product interest behavior, including the person to be predicted selected in step S321 and the measure to be predicted selected in step S320.

[0103] Then, in step S331, the data acquisition unit 400 refers to the person database 20 and the external environment database 22 based on the prediction conditions (the person to be predicted and the measure to be predicted), and acquires input data including person data, behavioral factor indicators, measure data, and external environment data for the prediction conditions.

[0104] Next, in step S332, the product interest behavior prediction unit 401 inputs the input data acquired in step S331 into the learning model 12, and predicts the person's product interest behavior based on the output data output from the learning model 12 as the inference result.

[0105] Next, in step S340, upon receiving from the policy extraction unit 403 the prediction results of product interest behavior for each of the policies and individuals targeted for prediction as a result of the loop processing in steps S320 and S321, the policy extraction unit 403 evaluates the effectiveness of each policy by statistically processing the product interest behavior of each individual for each of the multiple policies.The policy extraction unit 403 then extracts, from the multiple policies, policies whose effectiveness meets, for example, a predetermined criterion.

[0106] Next, in step S350, the output processing unit 404 transmits screen information for displaying the measures that meet the predetermined criteria on the display screen as the extraction results in step S341 to the policy planner terminal device 6. At this time, the output processing unit 404 may include in the screen information not only the extracted results of the measures, but also the effects of each measure and the predicted results of product interest behavior for each person.

[0107] Next, in step S360, upon receiving the screen information, the policy planner terminal device 6 displays on the display screen, based on the screen information, the results of extracting policies whose effects meet a predetermined standard from among the multiple policies to be predicted. This completes the series of processes shown in Fig. 13. In the data processing method, step S331 corresponds to a data acquisition step, step S332 corresponds to a product interest behavior prediction step, steps S310 to S340 correspond to a policy extraction step, and step S350 corresponds to an output processing step.

[0108] As described above, the data processing device 4 and data processing method according to this embodiment input data, including person data, behavioral factor indicators, policy data, and external environment data, into the trained learning model 12, and based on the output data output from the learning model 12, predict (infer) a person's product interest behavior regarding a product as a behavior caused by the policy. Therefore, a person's product interest behavior can be accurately predicted while taking into account the factors that motivate the person to take action and the impact of the policy implemented regarding the product on the person's emotions. Furthermore, the predicted results of each person's product interest behavior can be statistically processed and used in various services (product planning, product development, production management, purchasing and order management, sales strategy formulation, marketing, advertising, etc.).

[0109] (Other embodiments) The present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the spirit and scope of the present invention, all of which are included in the technical concept of the present invention.

[0110] In the above embodiment, the database device 2, the machine learning device 3, and the data processing device 4 are described as being configured as separate devices, but these three devices may be configured as a single device, or any two of these three devices may be configured as a single device. Furthermore, at least one of the machine learning device 3 and the data processing device 4 may be incorporated into the worker terminal device 5 or the policy planner terminal device 6.

[0111] In the above embodiment, a case has been described in which a neural network is used as a learning model for realizing machine learning by the machine learning unit 301, but other machine learning models may also be used. Examples of other machine learning models include tree models such as decision trees and regression trees, ensemble learning models such as bagging and boosting, and neural network models (deep neural networks) such as recurrent neural networks, convolutional neural networks, and LSTM. learning), hierarchical clustering, non-hierarchical clustering, k-nearest neighbors, k-means Examples of such methods include clustering models such as the average method, principal component analysis, factor analysis, multivariate analysis models such as logistic regression, and support vector machines.

[0112] In the above embodiment, the learning data 11 constitutes the input data input to the learning model 12 and includes person data, behavioral factor indexes, policy data, and external environment data. However, the external environment data may be omitted from the input data. That is, the input data may include person data, behavioral factor indexes, and policy data. In this case, the machine learning unit 301 of the machine learning device 3 inputs multiple sets of learning data 11, each consisting of input data (person data, behavioral factor indexes, and policy data) and output data (product interest behavior data), to the learning model 12, thereby causing the learning model 12 to learn the correlation between the input data and the output data through machine learning. Furthermore, the data acquisition unit 400 of the data processing device 4 acquires input data including person data, behavioral factor indexes, and policy data. The product interest behavior prediction unit 401 inputs the input data (person data, behavioral factor indexes, and policy data) acquired by the data acquisition unit 400 to the trained learning model 12, thereby predicting a person's product interest behavior based on the output data output from the learning model 12.

[0113] In the above embodiment, the machine learning device 3 and the data processing device 4 handle various data in units of processing, with a single person as the person to be predicted and learned. However, a group of multiple people may be considered as a single person, and the data processing method and the machine learning method may be performed with the group of multiple people as the person to be predicted and learned. In this case, for example, the person data included in the input data may be a collection or aggregated value of person data for each of the multiple people, the behavioral factor index included in the input data may be assigned based on the collection or aggregated value of person data for each of the multiple people, and the policy data may indicate a policy to be implemented for the multiple people. Furthermore, the product interest behavior data included in the output data may be a collection or aggregated value of product interest behavior for each of the multiple people (e.g., the total number or frequency of purchases made by multiple people, the proportion of people who have made a purchase among the multiple people, etc.).

[0114] As described above, by regarding a group of people consisting of multiple people as one person, the machine learning device 3 and the machine learning method can provide a learning model 12 that can predict (infer) the product interest behavior of multiple people (group of people) based on input data related to the multiple people (group of people). By inputting input data about a person (group of people) into a trained learning model 12, it is possible to predict (infer) the product interest behavior of the multiple people (group of people) based on the output data output from the learning model 12.

[0115] (Inference device, inference method or inference program) The present invention can be provided not only in the form of the data processing device 4 (data processing method or data processing program) according to the above embodiment, but also in the form of an inference device (inference method or inference program). In this case, the inference device (inference method or inference program) can include a memory and a processor, and the processor executes a series of processes. The series of processes includes a data acquisition process (data acquisition step) for acquiring input data, and an inference process (inference step) for inferring output data from the input data once the input data has been acquired in the data acquisition process.

[0116] By providing it in the form of an inference device (inference method or inference program), it can be applied to various devices more easily than when it is implemented in the data processing device 4. It will be naturally understood by those skilled in the art that when the inference device (inference method or inference program) infers a behavioral factor index, the inference method implemented by the index assignment unit may be applied using a trained learning model generated by the machine learning device 3 and machine learning method according to the above embodiment. [Explanation of symbols]

[0117] 1...Policy planning system, 2...Database device, 3...Machine learning device, 4...data processing device, 5...worker terminal device, 6...policy planner terminal device, 7...Network, 10A...Worker, 10B...Policy planner, 11...Learning data, 12...Learning model, 20...People database, 21...Policy database, 22...External environment database, 30...control unit, 31...storage unit, 32...communication unit, 33...input unit, 34...display unit, 40...control unit, 41...storage unit, 42...communication unit, 43...input unit, 44...display unit, 120...input layer, 121...intermediate layer, 122...output layer, 300...learning data acquisition unit, 301...machine learning unit, 310...machine learning program, 400...data acquisition unit, 401...product interest behavior prediction unit, 402...policy evaluation unit, 403...Measure extraction unit, 404...Output processing unit, 410...Data processing program

Claims

1. a data acquisition unit that acquires prediction input data including person data that records the behavior of a person to be predicted based on one or more perspectives, behavioral factor indicators that indicate factors that motivate the person to take the behavior, policy data that indicates a policy to be predicted that will be implemented for the person regarding a product, and external environment data that indicates the external environment at the time the policy is implemented; a product interest behavior prediction unit that predicts the product interest behavior of the person to be predicted based on the output data output from a learning model trained by machine learning to obtain a correlation between the prediction input data acquired by the data acquisition unit, the learning input data including person data recording the behavior of the person to be learned based on the viewpoint, behavior factor indexes indicating factors that motivate the person to take the behavior, policy data indicating policies to be implemented for the person to be learned regarding a product, and external environment data indicating the external environment at the time the policies are implemented, and output data including product interest behavior data recording product interest behavior with respect to the product as the behavior caused by the person as a result of the policies; Data processing device.

2. the system further comprises a policy evaluation unit that receives a person group including the plurality of persons to be predicted and a policy to be predicted, and sequentially selects the persons to be predicted from the person group, and the data acquisition unit acquires input data for prediction based on the persons to be predicted and the policy to be predicted, and performs a loop process that repeats the steps of: a data acquisition step in which the data acquisition unit acquires input data for prediction based on the persons to be predicted and the policy to be predicted; and a product interest behavior prediction step in which the product interest behavior prediction unit predicts the product interest behavior of the persons to be predicted based on the output data output from the learning model by inputting the input data for prediction selected by the data acquisition unit into the learning model; and receives, as a result of the loop process, a prediction result that predicts the product interest behavior of each of the plurality of persons included in the person group from the product interest behavior prediction unit, and evaluates the effectiveness of the policy by statistically processing the product interest behavior of each of the persons.

2. The data processing device according to claim 1.

3. A person group including the plurality of people to be predicted and a policy group including the plurality of policies to be predicted are received, and policies to be predicted are sequentially selected from the policy group, and people to be predicted are sequentially selected from the person group, and the data acquisition unit receives the people to be predicted and the policies to be predicted from the person group. and a product interest behavior prediction step in which the product interest behavior prediction unit inputs the input data for prediction selected by the data acquisition unit into the learning model, and predicts the product interest behavior of the person to be predicted based on the output data output from the learning model, and a policy extraction unit receives, as a result of the loop processing, prediction results of the product interest behavior of each of the plurality of persons included in the person group for each of the plurality of policies included in the policy group from the product interest behavior prediction unit, and extracts the policy that satisfies a predetermined criterion from the plurality of policies by statistically processing the product interest behavior of each person for each policy.

2. The data processing device according to claim 1.

4. a machine learning unit that inputs multiple sets of learning data into a learning model, the learning data including: person data that records the behavior of a person to be learned based on one or more perspectives; behavioral factor indicators that indicate factors that motivate the person to take the behavior; policy data that indicates the policy to be learned that is implemented for the person regarding a product; and external environment data that indicates the external environment at the time the policy is implemented; and output data that includes product interest behavior data that records product interest behavior regarding the product as the behavior caused by the person as a result of the policy; and causes the learning model to learn a correlation between the learning input data and the output data through machine learning; a learned model storage unit that stores the learned model in which the correlation is learned by the machine learning unit, Machine learning device.

5. A computer-implemented data processing method, comprising: a data acquisition step of acquiring input data for prediction, including person data recording the behavior of the person to be predicted based on one or more viewpoints, behavioral factor indicators indicating factors that motivate the person to take the behavior, policy data indicating policies to be implemented for the person regarding a product, and external environment data indicating the external environment at the time the policies are implemented; a product interest behavior prediction step of predicting the product interest behavior of the person to be predicted based on the output data output from a learning model trained by machine learning to obtain correlations between the prediction input data acquired by the data acquisition step, the learning input data including person data recording the behavior of the person to be learned based on the viewpoint, behavior factor indexes indicating factors that motivate the person to take the behavior, policy data indicating policies to be learned regarding a product to be implemented for the person, and external environment data indicating the external environment at the time the policies are implemented, and output data including product interest behavior data recording product interest behavior regarding the product as the behavior caused by the person as a result of the policies, Data processing methods.

6. A computer-implemented machine learning method, comprising: a machine learning process of inputting multiple sets of learning data into a learning model, the learning data including: person data recording the behavior of a person to be learned based on one or more perspectives; behavioral factor indicators indicating factors that motivate the person to take the behavior; policy data indicating the policy to be learned regarding a product to be implemented for the person; and external environment data indicating the external environment at the time the policy is implemented; and output data including product interest behavior data recording product interest behavior regarding the product as the behavior caused by the person as a result of the policy; and causing the learning model to learn the correlation between the learning input data and the output data through machine learning; and a learned model storage step of storing the learned model, which has learned the correlation through the machine learning step, in a learned model storage unit. Machine learning methods.

Images