Learning device, learning method, and learning program

Abstract

To make use of a vast amount of data accumulated in relation to use of services.SOLUTION: An information processing device according to the present invention includes an acquisition unit and a learning unit. Upon receiving user conditions related to a user as a target of learning, the acquisition unit acquires feature values indicating features of users who satisfy the conditions from a database storing feature amounts indicating features of users, and acquires action information indicating actions took by the users who satisfy the user conditions from a database storing action information indicating actions took by users. The learning unit makes a model learn a relation between the feature values and the action information acquired by the acquisition unit.SELECTED DRAWING: Figure 6

Description

[Technical field]

[0001] The present application relates to a learning device, a learning method, and a learning program. [Background technology]

[0002] Conventionally, various online services have been provided via the Internet, and huge amounts of data accumulated with the use of these services are analyzed. For example, a technology has been proposed that analyzes customers by taking into account trends over time using purchase history data, etc. (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2015-146145 A Summary of the Invention [Problem to be solved by the invention]

[0004] However, there is still room for improvement in conventional technologies. For example, advanced skills may be required to find and process useful data from the huge amount of data accumulated through the use of services, which may hinder the promotion of data utilization.

[0005] The present application has been made in consideration of the above, and aims to provide a learning device, a learning method, and a learning program that can utilize the vast amount of data accumulated as a result of service use. [Means for solving the problem]

[0006] The information processing device according to the present application includes an acquisition unit and a learning unit. When the acquisition unit receives a user condition related to a user to be learned, the acquisition unit acquires feature amounts indicating characteristics of a user who satisfies the user condition from a database in which feature amounts indicating characteristics of each user are registered, and acquires behavioral information indicating behavior performed by the user who satisfies the user condition from a database in which behavioral information indicating behavior performed by each user is registered. The learning unit trains a model on the relationship between the feature amounts and the behavioral information acquired by the acquisition unit. Effect of the Invention

[0007] According to one aspect of the embodiment, it is possible to utilize the vast amount of data accumulated as the service is used. [Brief description of the drawings]

[0008] [Figure 1] FIG. 1 is a diagram illustrating an example of the configuration of an information processing system according to an embodiment. [Diagram 2] FIG. 2 is a diagram for explaining an overview of the learning process (part 1) according to the embodiment. [Diagram 3] FIG. 3 is a diagram for explaining an overview of the learning process (part 2) according to the embodiment. [Figure 4] FIG. 4 is a diagram for explaining an overview of the score reproduction process according to the embodiment. [Diagram 5] FIG. 5 is a diagram illustrating an example of hierarchical information according to the embodiment. [Figure 6] FIG. 6 is a diagram illustrating an example of the configuration of a model creating unit according to the embodiment. [Figure 7] FIG. 7 is a diagram illustrating an overview of behavior information stored in the user behavior DB according to the embodiment. [Figure 8] FIG. 8 is a diagram illustrating an overview of information related to feature amounts stored in a user feature amount DB according to the embodiment. [Figure 9] FIG. 9 is a diagram illustrating an overview of information on a trained model stored in a model DB according to the embodiment. [Figure 10]FIG. 10 is a diagram showing an example of the configuration of a score reproduction unit according to the embodiment. [Figure 11] FIG. 11 is a flowchart illustrating an example of a learning process procedure (part 1) executed by the model creating unit according to the embodiment. [Figure 12] FIG. 12 is a flowchart illustrating an example of a procedure (part 2) of the learning process executed by the model creating unit according to the embodiment. [Figure 13] FIG. 13 is a flowchart illustrating an example of a procedure of the score reproduction process executed by the score reproduction unit according to the embodiment. [Figure 14] FIG. 14 is a hardware configuration diagram showing an example of a computer that realizes the functions of the model creating unit according to the embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] Below, a detailed description will be given of the form (hereinafter referred to as "embodiment") for implementing the learning device, learning method, and learning program according to the present application with reference to the drawings. Note that the learning device, learning method, and learning program according to the present application are not limited to this embodiment. Furthermore, each embodiment can be appropriately combined to the extent that there is no contradiction in the processing contents. Furthermore, the same parts in each of the following embodiments are given the same symbols, and duplicated explanations will be omitted.

[0010] [Embodiment] [1. Overview of information processing according to the embodiment] (1-1. System Configuration) An overview of the information processing system SYS for executing information processing according to the embodiment will be described below. Fig. 1 is a diagram showing an example of the configuration of the information processing system SYS according to the embodiment. The information processing system SYS shown in Fig. 1 has a model creation unit 100 which is an example of a learning device according to the present application.

[0011] 1, the information processing system SYS according to the embodiment includes a model creation unit 100 and a score reproduction unit 200. The model creation unit 100 and the score reproduction unit 200 are connected to a predetermined network N such as the Internet by wire or wirelessly. The model creation unit 100 and the score reproduction unit 200 can communicate with other devices such as the operator device 10 through the predetermined network N.

[0012] The model creation unit 100 executes, as information processing according to the embodiment, a process of creating a model for calculating a predicted score related to the service usage of a user who uses various online services. The model creation unit 100 is typically a server device, but may be realized by a mainframe, a workstation, or the like. In addition, when the model creation unit 100 is realized by a server device, it may be realized by a single server device, or may be realized by a cloud system in which multiple server devices and multiple storage devices operate in cooperation with each other.

[0013] The model creating unit 100 includes a user behavior DB 121 , a user feature DB 122 , and a model DB 123 .

[0014] The user behavior DB121 stores behavior information (i.e., behavior logs) related to behaviors performed by each user who is a service user using various online services. For example, the behavior information may include a visit record of an online service and a conversion record in an online service. The various online services may include an Internet connection, a search service, a travel information service, a social networking service (SNS), an electronic commerce service, an electronic payment service, an online game, an online banking service, an online trading service, a hotel reservation service, a ticket reservation service, a video distribution service, a music distribution service, a news distribution service, a map information service, a route search service, a route guidance service, a route information service, an operation information service, and a weather information service.

[0015] The user feature DB 122 stores information on features indicating the attributes, interests, and other characteristics of the above-mentioned service users. For example, the information on features stored in the user feature DB 122 is collected as learning data from various information recorded in various online services in order to learn a model for calculating a score related to the above-mentioned service user's use of the service. The information on features includes information on demographic attributes such as the user's age, sex, address, and family structure, and information on psychographic attributes such as interests and preferences.

[0016] The model DB 123 stores information about a trained model created by the model creation unit 100. The information about the trained model stored in the model DB 123 is created by having a model learn the relationship between the feature amount and behavioral information associated with the above-mentioned service user.

[0017] The score reproduction unit 200 reproduces a score that matches a prediction condition received from an operator (for example, the operator OP shown in FIGS. 2 to 4) who uses the information processing system SYS.

[0018] The score reproduction unit 200 has a model DB 221. The model DB 221 stores information of the trained model created by the score reproduction unit 100. The model DB 221 stores a plurality of models for calculating scores related to the behavior of the service user in various online services when information associated with the above-mentioned service user who uses various online services (an example of a "predetermined service") is input.

[0019] The operator device 10 is an information processing terminal used by an operator OP who uses the information processing system SYS. For example, the operator device 10 can be realized by a smartphone, a desktop PC (Personal Computer), a notebook PC, a tablet terminal, a PDA (Personal Digital Assistant), or the like. The operator device 10 can communicate with a model creation unit 100 and a score reproduction unit 200 included in the information processing system SYS through a network N. The operator device 10 may have a communication function for performing short-distance wireless communication such as Bluetooth (registered trademark) or a wireless LAN (Local Area Network).

[0020] (1-2. Learning process (part 1)) Hereinafter, an overview of the learning process (part 1) according to the embodiment will be described with reference to Fig. 2. Fig. 2 is a diagram for explaining an overview of the learning process (part 1) according to the embodiment. In the following description, an operator OP may be regarded as the same as the operator device 10. In other words, the operator OP may be rephrased as the operator device 10.

[0021] An operator OP accesses the model creating unit 100 and specifies a query to the user behavior DB 121 as a behavior condition, which is a model creating condition (step S01).

[0022] When the model creation unit 100 accepts a query specified by the operator OP (that is, when it receives a query from the operator device 10), it acquires a log of behavior matching the specified query from the user behavior DB 121 (step S02).

[0023] In addition, the model creation unit 100 acquires features of a relevant user who is a service user corresponding to the acquired behavior log, and features of a non-relevant user who is a service user who does not correspond to the acquired behavior log, from the user feature DB 122 (step S03).

[0024] Furthermore, the model creation unit 100 learns the model so as to output a high score when the feature amount of a relevant user is input, and output a low score when the feature amount of a non-relevant user is input (step S04).

[0025] For example, the model creation unit 100 may learn a model using a known machine learning method such as GBDT (Gradient Boosting Decision Tree), which is one of supervised machine learning methods, to create a trained model. GBDT includes XGBoost (eXtreme Gradient Boosting) and LightGBM (Gradient Boosting Machine). Note that the creation of a trained model includes not only the creation of a new trained model, but also the updating of an existing trained model. The model creation unit 100 registers information of the created trained model in the model DB 123.

[0026] (1-3. Learning process (part 2)) Hereinafter, an overview of the learning process (part 2) according to the embodiment will be described with reference to Fig. 3. Fig. 3 is a diagram for explaining an overview of the learning process (part 2) according to the embodiment. In the following description, the operator OP may be regarded as the operator device 10. In other words, the operator OP may be referred to as the operator device 10.

[0027] An operator OP accesses the model creating unit 100 and specifies feature quantities that are to be model creating conditions (step S11).

[0028] When the model creation unit 100 receives the feature quantity specified by the operator OP (i.e., when the feature quantity is received from the operator device 10), the model creation unit 100 identifies, from the user feature quantity DB 122, relevant users having feature quantities similar to the specified feature quantity and non-relevant users having no feature quantities similar to the specified feature quantity (step S12). Note that similar feature quantities include matching feature quantities.

[0029] Furthermore, the model creation unit 100 acquires a log of the identified user's behavior from the user behavior DB 121 (step S13).

[0030] In addition, the model creation unit 100 learns the model so that a high score is output when the feature of the corresponding user is input, and a low score is output when the feature of a non-corresponding user is input (step S14). The model creation unit 100 registers information about the created model in the model DB 123.

[0031] The model creation unit 100 may receive a model learning method designation from an operator OP as a model creation condition. The model creation unit 100 may learn a model at a predetermined timing, such as daily, and update the model information stored in the model DB 123.

[0032] In the above-mentioned learning process (part 1) and learning process (part 2), the model creation unit 100 estimates the importance of the behavior log and feature amount (hereinafter, "learning data") used to create the model for each type of data using a publicly known technique such as XGBoost, and extracts types of learning data with high importance (for example, gender, purchase date and time of a product, etc.) based on the estimated importance. Then, the model creation unit 100 re-learns the model using only the extracted learning data. The model creation unit 100 may also use a bandit algorithm to include new types of learning data as additional types of learning data in the learning target. The model creation unit 100 repeats this process to improve the accuracy of the model.

[0033] In the above-mentioned learning process (part 1) and learning process (part 2), the model creation unit 100 may learn a model for each channel through which information flows into the online service, using learning data corresponding to each channel. The channels may include e-mail, advertisements, service domains, and the like. In the above-mentioned learning process (part 1) and learning process (part 2), the model creation unit 100 may learn a model for each piece of hierarchical information that classifies the transition of behaviors up to a predetermined behavior in the online service used by the user into hierarchical stages, using learning data corresponding to each piece of hierarchical information.

[0034] (1-4. Score reproduction process) Hereinafter, an overview of the score reproduction process according to the embodiment will be described with reference to Fig. 4. Fig. 4 is a diagram for explaining the overview of the score reproduction process according to the embodiment. In the following description, the operator OP may be regarded as the operator device 10. In other words, the operator OP may be referred to as the operator device 10.

[0035] The operator OP accesses the score reproduction unit 200 and specifies a prediction condition for which the score is to be calculated (step S21). For example, a prediction condition (an example of a "new condition") specified by the operator OP is "a user who is likely to purchase product X".

[0036] When the score reproduction unit 200 receives the prediction conditions designated by the operator OP (i.e., when it receives them from the operator device 10), it selects a model that matches the received prediction conditions (step S22). For example, the score reproduction unit 200 selects each model for calculating a score used to reproduce a score indicating the degree of satisfaction of each condition included in the prediction conditions from among the multiple models based on a comparison result between the contents of the prediction conditions and the conditions corresponding to each of the multiple models stored in the model DB 221. For example, the score reproduction unit 200 can extract each condition included in the prediction conditions by natural language processing, and automatically extract a model corresponding to each extracted condition according to a rule base that is specified in advance. Note that the score reproduction unit 200 may extract each condition included in the prediction conditions by natural language processing, present the model corresponding to the extracted condition to the operator OP, and select the model selected by the operator OP.

[0037] Also, according to Figure 4, an example is shown in which the score reproduction unit 200 receives a prediction condition such as "users likely to purchase product X" from an operator OP, extracts "product X" and "purchase" as conditions included in the prediction condition, and selects an interest model corresponding to "product X" and a CV prediction model corresponding to "purchase".

[0038] The interest model inputs behavioral information associated with a user who is a service user, and outputs an interest score that indicates interest for each product or service. The conversion prediction model inputs features associated with a user who is a service user, and outputs a conversion prediction score that predicts with probability whether the user will purchase a product within a certain period of time.

[0039] Furthermore, the score reproduction unit 200 reproduces the score that matches the prediction condition by using the scores calculated by each selected model (step S23).

[0040] For example, when the score calculated by each selected model is a probability, the score reproduction unit 200 multiplies a predicted score calculated for each user by a first model among the selected models with a predicted score calculated for each user by a second model among the selected models to reproduce a score indicating the likelihood that the user will perform the behavior to be estimated.

[0041] For example, when it is difficult to multiply the scores calculated by the selected models, the score reproduction unit 200 reproduces a score for predicting a user who is likely to perform a behavior to be estimated by using the ranking of the predicted scores calculated for each user by a first model among the selected models and the ranking of the predicted scores calculated for each user by a second model among the selected models. For example, in the case shown in Fig. 4, the score reproduction unit 200 calculates, as a reproduction score, a combination of the ranking of the CV prediction scores output by the CV prediction model for each user who is a service user and the ranking of the interest scores output by the interest model for each user who is a service user.

[0042] In addition, the score reproduction unit 200 may receive, as a prediction condition, a specification of information on a channel that flows into the online service when the user uses the online service. The channel may include e-mail, advertisement, service domain, etc. In this case, the score reproduction unit 200 selects a model corresponding to the channel specified as the prediction condition and performs score reproduction.

[0043] Furthermore, the score reproduction unit 200 may receive, as a prediction condition, an estimate of hierarchical information in which the transition of behaviors up to a predetermined behavior in an online service used by a user is hierarchically divided into stages. FIG. 5 is a diagram showing an example of hierarchical information according to an embodiment. As shown in FIG. 5, the hierarchical information may include no visit (new), awareness, visit, use, app download (DL), continuation, and the like. In this case, the score reproduction unit 200 selects a model corresponding to a funnel specified as a prediction condition, and executes reproduction of the score.

[0044] Furthermore, when the score reproduction unit 200 does not specify each model for calculating the score used to reproduce the score indicating the degree of satisfaction of each condition included in the prediction condition, the score reproduction unit 200 may transmit a learning request to the model creation unit 100 to make the model learn the relationship between the conditions included in the prediction condition. Then, the score reproduction unit 200 may register the information of the model received from the model creation unit 100 in the model DB 221. [2. Equipment configuration] (2-1. Model Creation Unit 100) An example of the functional configuration of each device included in the information processing system SYS according to the embodiment will be described below. First, an example of the functional configuration of the model creation unit 100 included in the information processing system SYS will be described. FIG. 6 is a diagram showing an example of the configuration of the model creation unit 100 according to the embodiment. As shown in FIG. 6, the model creation unit 100 has a communication unit 110, a storage unit 120, and a control unit 130.

[0045] (Regarding communication unit 110) The communication unit 110 is realized by, for example, a network interface card (NIC). The communication unit 110 is connected to a network N by wire or wirelessly. The model creation unit 100 transmits and receives information to and from other devices such as the operator device 10 and the score reproduction unit 200 via the network N.

[0046] (Regarding the storage unit 120) The storage unit 120 stores, for example, programs and data used for control and calculation by the control unit 130. For example, the storage unit 120 is realized by a semiconductor memory element such as a random access memory (RAM) or a flash memory, or a storage device such as a hard disk or an optical disk. For example, the storage unit 120 has a user behavior DB 121, a user feature DB 122, and a model DB 123. Note that the storage unit 120 is not limited to the example shown in FIG. 6, and can store data required for executing the information processing according to the embodiment as appropriate.

[0047] (User behavior DB121) The user behavior DB 121 stores behavior information (i.e., behavior logs) related to the behavior of each user who is a service user of various online services. Fig. 7 is a diagram showing an overview of the behavior information stored in the user behavior DB 121 according to the embodiment.

[0048] 7, the behavior information stored in the user behavior DB 121 has an item of "user ID" and an item of "behavior information." These items of the behavior information are associated with each other.

[0049] The "user ID" field stores a user ID, which is identification information specific to each user and is assigned to each user to uniquely identify each user who is a service user of various online services.

[0050] In the "behavioral information" item, for example, for various online services, behavior logs such as visit records of online services, such as PV (Page View), and conversion records in online services, such as CV, are stored.

[0051] (User feature DB122) The user feature DB 122 stores information on features indicating characteristics such as attributes and interests of each user who is a service user of various online services. Fig. 8 is a diagram showing an overview of information on features stored in the user feature DB 122 according to the embodiment.

[0052] 8, the information on the feature amount stored in the user feature amount DB 122 has an item of "user ID" and an item of "feature amount." These items of the information on the feature amount are associated with each other.

[0053] The "user ID" field stores a user ID, which is identification information unique to each user that is assigned to each user in order to uniquely identify each user who is a service user of various online services. For example, this user ID is the same information as the information stored in the "user ID" field of the user behavior DB 121.

[0054] The "feature amount" item stores information about feature amounts indicating characteristics such as the attributes and interests of each user. The feature amount information is collected as learning data from various information recorded in various online services in order to learn a model for calculating a score related to the service usage of a service user. The feature amount information includes information on demographic attributes such as the user's age, sex, address, and family structure, and information on psychographic attributes such as interests and preferences. The feature amount information may be data obtained by estimating and scoring the attributes and interests of each user.

[0055] (Model DB123) The model DB 123 stores information about trained models created by the model creating unit 100. Fig. 9 is a diagram illustrating an overview of information about trained models stored in the model DB 123 according to the embodiment.

[0056] 9, the trained model information stored in the model DB 123 has an item of “model ID”, an item of “corresponding condition”, and an item of “model information”. These items in the trained model information are associated with each other.

[0057] The "model ID" field stores a model ID, which is identification information specific to each model and is assigned to each model in order to uniquely identify the model.

[0058] The "Corresponding Condition" item stores information indicating conditions corresponding to a model. For example, the conditions corresponding to a model include visit prediction, conversion prediction, and interest estimation. In addition, detailed conditions such as travel site visit prediction and conversion prediction via advertisement may be added to the conditions corresponding to a model. In addition, information on channels and hierarchical information (also called funnels) may be added to the conditions corresponding to a model.

[0059] The "model information" item stores information about the trained model. For example, when machine learning is performed using a neural network, the trained model information may include various information such as connection information about how the nodes included in each of the multiple layers that make up the neural network are connected to each other, and connection coefficients that are multiplied by the numerical values ​​input and output between the connected nodes.

[0060] (Regarding the control unit 130) The control unit 130 is realized by a CPU (Central Processing Unit), an MPU (Micro Processing Unit), or the like executing various programs stored in a storage device inside the model creation unit 100 using a RAM as a working area. The control unit 130 may also be realized by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

[0061] The control unit 130 shown in Fig. 6 has an acquisition unit 131 and a learning unit 132, and these units realize or execute the functions and actions of the information processing described below. The control unit 130 may have a plurality of internal configurations divided into processing units that realize or execute the functions and actions of the information processing described below. The control unit 130 is not limited to the configuration shown in Fig. 6, and may have other configurations as long as they perform the information processing described below, and may have other functional units other than those shown in Fig. 6.

[0062] (Acquisition part 131) When the acquisition unit 131 receives user conditions related to a user to be learned, the acquisition unit 131 acquires feature quantities indicating the characteristics of a user who satisfies the user conditions from a user feature quantity DB 122, which is a database in which feature quantities indicating the characteristics of each user are registered, and acquires behavior information indicating the behavior performed by the user who satisfies the user conditions from a user behavior DB 121, which is a database in which behavior information indicating the behavior performed by each user is registered. The acquisition unit 131 passes the acquired feature quantities and behavior information to the learning unit 132.

[0063] In addition, when the acquisition unit 131 receives a query specifying the behavior of a user to be learned from an operator OP, who is a user of the model creation unit 100, it acquires behavior information corresponding to the behavior specified by the query and acquires features associated with the acquired behavior information.

[0064] In addition, when the acquisition unit 131 receives a specification of a user feature to be learned from an operator OP, who is a user of the model creation unit 100, it acquires a feature amount corresponding to the specified feature and acquires behavioral information linked to the acquired feature amount.

[0065] The behavioral information acquired by the acquisition unit 131 may include information on a channel through which a user to be studied enters a predetermined service when the user uses the predetermined service. The behavioral information acquired by the acquisition unit 131 may include hierarchical information that classifies, in a step-by-step manner, a transition of behavior leading to a predetermined behavior in a predetermined service used by the user to be studied.

[0066] (Learning Section 132) The learning unit 132 trains a model to learn the relationship between the feature amount acquired by the acquisition unit 131 and the behavioral information. The learning unit 132 can train the model using a known machine learning technique. For example, the learning unit 132 may train the model by supervised learning in which the behavioral information stored in the user behavior DB 121 is used as correct answer data and the feature amount stored in the user feature amount DB 122 is used as learning feature amount.

[0067] The machine learning method used by the learning unit 132 is not particularly limited, but may be learning using a GBDT (Gradient Boosting Decision Tree) such as LightGBM, or preparing learning data that links data (input information) with its correct answer information (output information), and learning the learning data using a multi-layer neural network such as a DNN (Deep Neural Network), etc.

[0068] (2-2. Score Reproduction Unit 200) An example of the functional configuration of the score reproduction unit 200 included in the information processing system SYS will be described below. Fig. 10 is a diagram showing an example of the configuration of the score reproduction unit 200 according to the embodiment. As shown in Fig. 10, the score reproduction unit 200 has a communication unit 210, a storage unit 220, and a control unit 230.

[0069] (Regarding communication unit 210) The communication unit 210 is realized by, for example, a network interface card (NIC). The communication unit 210 is connected to a network N by wire or wirelessly. The score reproduction unit 200 transmits and receives information to and from other devices such as the operator device 10 and the model creation unit 100 via the network N.

[0070] (Regarding the storage unit 220) The storage unit 220 stores, for example, programs and data used for control and calculation by the control unit 230. For example, the storage unit 220 is realized by a semiconductor memory element such as a random access memory (RAM) or a flash memory, or a storage device such as a hard disk or an optical disk. For example, the storage unit 220 has a model DB 221. Note that the storage unit 220 is not limited to the example shown in FIG. 10 and can store data required for executing the information processing according to the embodiment as appropriate.

[0071] (Model DB221) The model DB 221 stores information about trained models created by the model creating unit 100. The information about trained models stored in the model DB 221 is acquired from the model creating unit 100.

[0072] (Regarding the control unit 230) The control unit 230 is realized by a CPU (Central Processing Unit) or an MPU (Micro Processing Unit) executing various programs stored in a storage device inside the score reproduction unit 200 using a RAM as a working area. The control unit 230 may also be realized by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

[0073] 10 has a receiving unit 231, a selecting unit 232, a reproducing unit 233, a learning request unit 234, and a registering unit 235, and these units realize or execute the functions and actions of the information processing described below. Note that the control unit 230 may have a plurality of internal configurations divided into processing units that realize or execute the functions and actions of the information processing described below. Also, the control unit 230 is not limited to the configuration shown in FIG. 10, and may have other configurations as long as they perform the information processing described below, and may have other functional units other than those shown in FIG. 10.

[0074] (Reception Department 231) The receiving unit 231 receives a designation of a new condition for which a score is to be calculated from an operator OP who is a user of the score reproduction unit 200. For example, the receiving unit 231 receives a designation of a prediction condition related to a user's behavior from the operator OP.

[0075] The receiving unit 231 may also receive, as a prediction condition, designation of information on a channel through which information flows into a service when the user uses various online services. The receiving unit 231 may also receive, as a prediction condition, an estimate of hierarchical information (also called a funnel) in which transitions of actions until a predetermined action is reached in various online services used by the user are hierarchically divided in stages.

[0076] (Selection Department 232) The selection unit 232 selects each model for calculating a score used to reproduce a score indicating the degree to which each condition included in the new condition is satisfied from among the multiple models, based on a comparison result between the content of the new condition accepted by the acceptance unit 231 and the condition corresponding to each of the multiple models stored in the model DB 221. For example, the selection unit 232 selects a first model and a second model as each model from among the multiple models, based on a comparison result between the content of the prediction condition accepted by the acceptance unit 231 and the condition corresponding to each of the multiple models.

[0077] (Recapitulation 233) The reproduction unit 233 reproduces a score indicating the degree to which a new condition is satisfied, using the score calculated by each model selected by the selection unit 232. For example, the reproduction unit 233 reproduces a score indicating the possibility that an action will be performed in various online services for each user by multiplying the score calculated by the first model for each user by the score calculated by the second model for each user. Also, for example, the reproduction unit 233 reproduces a score for predicting a user who is likely to perform an action in various online services, using the ranking of the predicted scores calculated by the first model for each user and the ranking of the predicted scores calculated by the second model for each user.

[0078] (Learning Request Section 234) If the selection unit 232 does not select each model for calculating the score to be used to reproduce the score indicating the degree to which each condition included in the new conditions is satisfied, the learning request unit 234 sends a learning request to the model creation unit 100 to have the model learn the relationships between the conditions included in the new conditions.

[0079] (Registration Section 235) The registration section 235 registers in the model DB 221 information about the model received from the model creation unit 100 (an example of an “external device”).

[0080] 3. Processing procedure according to the embodiment (3-1. Learning process procedure (part 1)) The procedure (part 1) of the learning process executed by the model creation unit 100 according to the embodiment will be described below. First, the procedure (part 1) of the learning process executed by the model creation unit 100 according to the embodiment will be described with reference to FIG. 11. FIG. 11 is a flowchart showing an example of the procedure (part 1) of the learning process executed by the model creation unit 100 according to the embodiment. The processing procedure shown in FIG. 11 is executed by the control unit 130 of the model creation unit 100. The processing procedure shown in FIG. 11 is repeatedly executed while the model creation unit 100 is operating.

[0081] 11, the receiving unit 131 receives a query specifying a user's behavior to be learned from an operator OP (step S101). For example, this query is a query to the user behavior DB 121.

[0082] When the learning unit 132 receives the query specified by the operator OP, it acquires a log of an action matching the specified query from the user action DB 121 (step S102).

[0083] In addition, the learning unit 132 acquires the features of the relevant user, who is a service user corresponding to the acquired behavior log, and the features of the non-relevant user, who is a service user not corresponding to the acquired behavior log, from the user feature DB 122 (step S103).

[0084] Furthermore, the learning unit 132 uses a predetermined model to learn the relationship between the behavioral information and the feature amount (step S104). For example, the learning unit 132 uses an arbitrary machine learning method to learn a model so as to output a high score when the feature amount of the corresponding user is input, and output a low score when the feature amount of a non-corresponding user is input.

[0085] In addition, the learning unit 132 uses publicly known technology to estimate the importance of the behavioral logs and features used to create the model for each type of data, and extracts types of behavioral logs or features with high importance (e.g., gender, product purchase date and time, etc.) based on the estimated importance (step S105).

[0086] Furthermore, the learning unit 132 re-learns the model using only the extracted learning data (step S106).

[0087] Furthermore, the learning unit 132 registers the re-learned model in the model DB 121 (step S107), and ends the procedure of the learning process shown in FIG.

[0088] (3-2. Learning process procedure (part 2)) The learning process procedure (part 2) executed by the model creation unit 100 according to the embodiment will be described below. First, the learning process procedure (part 2) executed by the model creation unit 100 according to the embodiment will be described with reference to FIG. 12. FIG. 12 is a flowchart showing an example of the learning process procedure (part 2) executed by the model creation unit 100 according to the embodiment. The process procedure shown in FIG. 12 is executed by the control unit 130 of the model creation unit 100. The process procedure shown in FIG. 12 is repeatedly executed while the model creation unit 100 is in operation. Note that the learning process procedure (part 2) shown in FIG. 12 differs from the learning process procedure (part 1) shown in FIG. 11 in steps S201 to S203.

[0089] As shown in FIG. 12, the receiving unit 131 receives, from an operator OP, a designation of a feature amount of a user to be a learning target (step S201).

[0090] When the learning unit 132 receives the features specified by the operator OP, it identifies relevant users having features similar to the specified features and non-relevant users who do not have features similar to the specified features from the user feature DB 122 (step S202).

[0091] Furthermore, the learning unit 132 acquires a log of the identified user's behavior from the user behavior DB 122 (step S203).

[0092] Furthermore, the learning unit 132 uses a predetermined model to learn the relationship between the behavioral information and the feature amount (step S204). For example, the learning unit 132 uses an arbitrary machine learning method to learn a model so as to output a high score when the feature amount of the corresponding user is input, and output a low score when the feature amount of a non-corresponding user is input.

[0093] In addition, the learning unit 132 uses publicly known technology to estimate the importance of the behavioral logs and features used to create the model for each type of data, and extracts types of behavioral logs or features with high importance (e.g., gender, product purchase date and time, etc.) based on the estimated importance (step S205).

[0094] Furthermore, the learning unit 132 re-learns the model using only the extracted learning data (step S206).

[0095] Furthermore, the learning unit 132 registers the re-learned model in the model DB 121 (step S107), and ends the procedure of the learning process shown in FIG.

[0096] (3-3. Score reproduction process procedure) The procedure of the score reproduction process executed by the score reproduction unit 200 according to the embodiment will be described below. First, the procedure of the score reproduction process executed by the score reproduction unit 200 according to the embodiment will be described with reference to FIG. 13. FIG. 13 is a flowchart showing an example of the procedure of the score reproduction process executed by the score reproduction unit 200 according to the embodiment. The processing procedure shown in FIG. 13 is executed by the control unit 230 of the score reproduction unit 200. The processing procedure shown in FIG. 13 is repeatedly executed while the score reproduction unit 200 is operating.

[0097] As shown in FIG. 13, the receiving unit 231 receives, from an operator OP, a specification of a prediction condition for which a score is to be calculated (step S301).

[0098] The selection unit 232 selects each model from among the multiple models based on the results of comparing the contents of the prediction conditions specified by the operator OP with the conditions corresponding to each of the multiple models stored in the model DB 221 to calculate a score used to reproduce the score indicating the degree to which each condition included in the prediction conditions is satisfied (step S302).

[0099] The reproduction unit 233 reproduces the score indicating the degree to which the new condition is satisfied, using the score calculated by each model selected by the selection unit 232 (step S303), and ends the procedure of the score reproduction process shown in FIG.

[0100] [4. Modifications] The information processing device, the information processing method, and the information processing program according to the present application may be embodied in various different forms other than the above-described embodiment. Modifications of the above-described embodiment will be described below.

[0101] In the above embodiment, an example has been described in which the information processing system SYS according to the embodiment is configured to include the model creation unit 100 and the score reproduction unit 200, but this example is not particularly limited. For example, the information processing system SYS may be configured to include a single information processing device in which the model creation unit 100 and the score reproduction unit 200 are functionally and physically integrated.

[0102] [5. Effects] The model creation unit 100 according to the embodiment includes an acquisition unit 131 and a learning unit 132. When the acquisition unit 131 receives a user condition related to a user to be learned, the acquisition unit 131 acquires a feature amount indicating the characteristics of a user who satisfies the condition from a user feature amount DB 122 in which a feature amount indicating the characteristics of each user is registered, and acquires behavior information indicating the behavior performed by the user who satisfies the user condition from a user behavior DB 121 in which behavior information indicating the behavior performed by each user is registered. The learning unit 132 causes a model to learn the relationship between the feature amount and the behavior information acquired by the acquisition unit 131.

[0103] In addition, when the acquisition unit 131 receives a query specifying the behavior of a user to be learned from an operator OP, who is a user of the model creation unit 100, the acquisition unit 131 acquires behavior information corresponding to the behavior specified by the query from the user behavior DB 121, and acquires features associated with the acquired behavior information from the user feature DB 122.

[0104] In addition, when the acquisition unit 131 receives a specification of a user feature to be learned from an operator OP, who is a user of the model creation unit 100, the acquisition unit 131 acquires a feature quantity corresponding to the specified feature from the user feature DB 122, and acquires behavioral information linked to the acquired feature quantity from the user behavior DB 121.

[0105] In this way, the model creation unit 100 according to the embodiment can utilize the data accumulated in the provision of various online services as learning data through the processes executed by each of the above-mentioned parts, or any combination of processes executed by each part, and learn a model that outputs a score related to the service usage for each service user, thereby making it possible to make effective use of the big data accumulated in various online services.

[0106] [6. Hardware configuration] Moreover, the model creation unit 100 according to the above-described embodiment and modified examples is realized, for example, by a computer 1000 having a configuration as shown in Fig. 14. Fig. 14 is a hardware configuration diagram showing an example of a computer that realizes the functions of the model creation unit according to the embodiment and modified examples.

[0107] The computer 1000 is connected to an output device 1010 and an input device 1020, and has a configuration in which an arithmetic unit 1030, a primary storage device 1040, a secondary storage device 1050, an output IF (Interface) 1060, an input IF 1070, and a network IF 1080 are connected via a bus 1090.

[0108] The arithmetic device 1030 operates based on programs stored in the primary storage device 1040 and the secondary storage device 1050 and programs read from the input device 1020, and executes various processes. The primary storage device 1040 is a memory device, such as a RAM, that temporarily stores data used by the arithmetic device 1030 for various calculations. The secondary storage device 1050 is a storage device in which data used by the arithmetic device 1030 for various calculations and various databases are registered, and is realized by a ROM (Read Only Memory), HDD, flash memory, etc.

[0109] The output IF 1060 is an interface for transmitting information to be output to an output device 1010 that outputs various types of information, such as a monitor or a printer, and is realized by a connector conforming to a standard such as USB (Universal Serial Bus), DVI (Digital Visual Interface), or HDMI (High Definition Multimedia Interface), a registered trademark. The input IF 1070 is an interface for receiving information from various input devices 1020, such as a mouse, keyboard, and scanner, and is realized by a USB, for example.

[0110] The input device 1020 may be a device that reads information from, for example, an optical recording medium such as a CD (Compact Disc), a DVD (Digital Versatile Disc), or a PD (Phase change rewritable Disk), a magneto-optical recording medium such as an MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory. The input device 1020 may also be an external storage medium such as a USB memory.

[0111] The network IF 1080 receives data from other devices via the network N and sends it to the arithmetic device 1030, and also transmits data generated by the arithmetic device 1030 to other devices via the network N.

[0112] The arithmetic unit 1030 controls the output device 1010 and the input device 1020 via the output IF 1060 and the input IF 1070. For example, the arithmetic unit 1030 loads a program from the input device 1020 or the secondary storage device 1050 onto the primary storage device 1040 and executes the loaded program.

[0113] For example, when the computer 1000 functions as the model creation unit 100 according to the embodiment or the modified example, the arithmetic device 1030 of the computer 1000 executes a program (e.g., a learning program) loaded onto the primary storage device 1040, thereby realizing a function similar to that of the control unit 130. That is, the arithmetic device 1030 cooperates with the program (e.g., a learning program) loaded onto the primary storage device 1040 to realize the processing by the model creation unit 100 according to the embodiment.

[0114] [7.Other] Of the processes described in the above embodiments, all or part of the processes described as being performed automatically can be performed manually, or all or part of the processes described as being performed manually can be performed automatically by a known method. In addition, the information including the process procedures, specific names, various data and parameters shown in the above documents and drawings can be changed arbitrarily unless otherwise specified.

[0115] In addition, each component of each device shown in the figure is a functional concept, and does not necessarily have to be physically configured as shown in the figure. In other words, the specific form of distribution and integration of each device is not limited to that shown in the figure, and all or part of them can be functionally or physically distributed and integrated in any unit according to various loads, usage conditions, etc.

[0116] Furthermore, the above-described embodiments can be appropriately combined as long as the processing contents are not contradictory.

[0117] The above describes the embodiments of the present application in detail with reference to several drawings. However, these are merely examples, and the present invention can be embodied in other forms that incorporate various modifications and improvements based on the knowledge of those skilled in the art, including the aspects described in the Disclosure of the Invention section.

[0118] Moreover, the above-mentioned "section, module, unit" can be read as "means" or "circuit", etc. For example, a control section can be read as control means or a control circuit. [Explanation of symbols]

[0119] SYS Information Processing System N Network 10 Operator device 100 model making units 110 Communications Department 120 Storage section 121 User Behavior DB 122 User feature DB 123 Model DB 130 Control section 131 Acquisition Department 132 Learning Department 200 Score Reproduction Unit 210 Communications Department 220 Storage section 221 Model DB 230 Control Unit 231 Reception 232 Selection Department 233 Recapitulation 234 Learning Request Department 235 Registration Department

Claims

1. an acquisition unit that, when receiving user conditions related to a user to be learned, acquires feature amounts indicating characteristics of a user who satisfies the user conditions from a database in which feature amounts indicating characteristics of each user are registered, and acquires behavior information indicating behavior performed by the user who satisfies the user conditions from a database in which behavior information indicating behavior performed by each user is registered; a learning unit that learns a relationship between the feature amount acquired by the acquisition unit and the behavioral information into a model; A learning device comprising:

2. The acquisition unit is When a query specifying a user behavior to be learned is received from a user, the behavior information corresponding to the behavior specified by the query is acquired, and the feature amount associated with the acquired behavior information is acquired.

2. The learning device according to claim 1 .

3. The acquisition unit is When a designation of a user characteristic to be learned is accepted from a user, the feature amount corresponding to the designated designation is acquired, and the behavior information associated with the acquired feature amount is acquired.

2. The learning device according to claim 1 .

4. 1. A computer implemented method of learning, comprising: an acquisition step of acquiring, upon receiving user conditions for a user to be learned, feature amounts indicating characteristics of a user who satisfies the user conditions from a database in which feature amounts indicating characteristics of each user are registered, and acquiring behavioral information indicating behavior performed by the user who satisfies the user conditions from a database in which behavioral information indicating behavior performed by each user is registered; a learning step of making a model learn the relationship between the feature amount acquired in the acquisition step and the behavioral information; A learning method comprising:

5. On the computer, an acquisition step of acquiring, when a user condition related to a user to be learned is received, a feature amount indicating the characteristics of a user who satisfies the user condition from a database in which a feature amount indicating the characteristics of each user is registered, and acquiring behavior information indicating the behavior performed by the user who satisfies the user condition from a database in which behavior information indicating the behavior performed by each user is registered; a learning step of making a model learn the relationship between the feature amount acquired by the acquisition step and the behavioral information; A learning program characterized by executing the above.

Images