Associated learning model generation device, associative learning model generation system, associative learning model generation method, program, and associative learning model

Abstract

Provided are a federated learning model and a device for generating the same, etc., that contribute simply and suitably to marketing activities. A federated learning model generating device (100) includes a local learning model acquiring unit (101) and a federated learning model generating unit (102). The local learning model acquiring unit (101) acquires a plurality of different local learning models that have learned a relationship between a plurality of customer groups generated from business operator customer data possessed by each of a plurality of business operators, and consumption behavior corresponding to the business operators. The federated learning model generating unit (102) accepts, as input data, a prescribed consumption behavior of a customer, by federating at least some of the acquired local learning models, and generates a federated learning model that outputs prospective customer data in response to the input data.

Description

[Technical Field]

[0001] The present invention relates to an associative learning model generation device, an associative learning model generation system, an associative learning model generation method, a program, and an associative learning model. [Background technology]

[0002] Businesses that provide products and services utilize various market data to gain business opportunities. Proposals have also been disclosed that utilize technologies such as machine learning as a method for utilizing market data.

[0003] Patent Document 1 discloses a technique for predicting a user's behavior using the user's behavior history in multiple domains.

[0004] Patent Document 2 discloses a technology that accepts from a user the purpose of use of information registered by a plurality of businesses, and presents the purpose of use to an analyst who analyzes the information according to the purpose of use.

[0005] Patent Document 3 discloses a technology in which an initial trained model that controls a specific operating device is incorporated into multiple operating devices, and multiple individual trained models obtained by additional learning based on individual operation data obtained by operating each of the operating devices are integrated and processed. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Japanese Patent Application Laid-Open No. 2016-076097 [Patent Document 2] Japanese Patent Application Publication No. 2019-046178 [Patent Document 3] Japanese Patent Application Publication No. 2020-038699 Summary of the Invention

[0007] If there is an abundance of good data that can serve as the basis for marketing regarding customer behavior in purchasing products or using services (i.e., specific consumer behavior), businesses will be able to more easily utilize this data for marketing. Therefore, methods for multiple businesses to use each other's customer data may be considered. However, it is not possible to share personal customer information across multiple businesses.

[0008] In view of the above-mentioned problems, an object of the present disclosure is to provide an associative learning model and a generation device thereof that contributes to marketing activities in a simple and suitable manner.

[0009] A federated learning model generation device according to one aspect of the present disclosure includes a local learning model acquisition unit and a federated learning model generation unit. The local learning model acquisition unit acquires multiple different local learning models that have learned the relationships between multiple customer groups generated from business customer data held by multiple businesses and the consumption behavior corresponding to the businesses. The federated learning model generation unit generates a federated learning model that accepts predetermined consumption behavior of customers as input data and outputs potential customer data for the input data by federating at least some of the acquired local learning models.

[0010] In the federated learning model generation method according to one aspect of the present disclosure, a computer executes the following processes: The computer acquires multiple different local learning models that have learned the relationship between multiple customer groups generated from business customer data held by multiple businesses and the consumption behavior corresponding to the businesses; The computer generates an federated learning model that accepts the predetermined consumption behavior of customers as input data and outputs potential customer data for the input data by federating at least some of the acquired local learning models.

[0011] A program according to one aspect of the present disclosure causes a computer to execute the following process: The computer acquires multiple different local learning models that have learned the relationship between multiple customer groups generated from business customer data held by multiple businesses and the consumption behavior corresponding to the businesses. The computer generates an federated learning model by federating at least some of the acquired local learning models, which accepts the predetermined consumption behavior of customers as input data and outputs potential customer data for the input data.

[0012] According to the present disclosure, it is possible to provide an associative learning model and a generation device thereof that contributes to marketing activities simply and suitably. [Brief explanation of the drawings]

[0013] [Figure 1] 1 is a block diagram of a data processing device according to a first embodiment; [Figure 2] 1 is a flowchart of a data processing method according to the first embodiment. [Figure 3] 1 is a block diagram of an associative learning model generation device according to a first embodiment. [Figure 4] 1 is a flowchart of an associative learning model generation method according to the first embodiment. [Figure 5] FIG. 10 is a block diagram of a data processing system according to a second embodiment. [Figure 6] FIG. 10 is a block diagram of a data processing device according to a second embodiment. [Figure 7] FIG. 10 is a block diagram of a local data processing device according to a second embodiment. [Figure 8] FIG. 10 is a diagram illustrating the processing of a business customer data generation unit. [Figure 9] FIG. 10 is a diagram showing business customer data. [Figure 10] FIG. 1 illustrates the processing of a local learning model. [Figure 11] FIG. 10 is a sequence diagram showing the processing executed by the federated learning model generation system. [Figure 12] FIG. 1 is a diagram illustrating a configuration of an associative learning model. [Figure 13] FIG. 1 illustrates the processing of a federated learning model. [Figure 14] FIG. 2 is a sequence diagram illustrating processing executed by the data processing system. [Figure 15] FIG. 10 is a block diagram of a data processing device according to a third embodiment. [Figure 16] FIG. 10 is a block diagram of a local data processing device according to a third embodiment. [Figure 17] FIG. 11 is a sequence diagram of an update method according to the third embodiment. [Figure 18] FIG. 2 is a block diagram illustrating an example of a hardware configuration of a computer. DETAILED DESCRIPTION OF THE INVENTION

[0014] The present invention will be described below through embodiments of the invention, but the invention according to the claims is not limited to the following embodiments. Furthermore, not all of the configurations described in the embodiments are necessarily essential means for solving the problems. For clarity of explanation, the following description and drawings have been omitted and simplified as appropriate. Note that in each drawing, the same elements are given the same reference numerals, and duplicate explanations are omitted as necessary.

[0015] <Embodiment 1> First, a first embodiment of the present disclosure will be described. Fig. 1 is a block diagram of a data processing device according to the first embodiment. A data processing device 110 shown in Fig. 1 receives input data related to a predetermined consumption behavior and outputs potential customer data for this consumption behavior.

[0016] The predetermined consumption behavior is, for example, the purchase of a product or ticket, or the use of a service, which directly or indirectly involves monetary payment. The potential customer data is data that defines, for example, customers who are relatively likely to engage in the above consumption behavior by a predetermined segment. Specifically, the potential customer data can be grouped by, for example, age, gender, address, family structure, hobbies, occupation, medical history, or past consumption behavior history. As a result, the data processing device 110 provides, for example, a predetermined business with data that can be used as a reference for determining target customer segments for marketing activities. Marketing activities are activities conducted by a business, and include sales, advertising, and sales or provision of products and services to customers.

[0017] The data processing device 110 may be configured, for example, by a computer, server, or dedicated device with communication capabilities. In the following description, the term "computer" may encompass server devices, blades, and cloud computing systems. The data processing device 110 mainly includes an input unit 111, a federated learning model 112, and an output unit 113.

[0018] The input unit 111 receives input data relating to a predetermined consumer behavior from a predetermined external device or the like. The input data relating to the predetermined consumer behavior is, for example, data indicating a predetermined product or service. The input data relating to the consumer behavior may also indicate detailed specifications of the product or service.

[0019] The federated learning model 112 is generated by federating at least some of the local learning models generated for multiple different operators. In this embodiment, "federation" refers to, for example, connecting the data structures of multiple local learning models, but the definition of federation is not limited to this. The federated learning model 112 is generated by federating multiple local learning models, thereby incorporating the characteristics of the local learning models. Furthermore, the federated learning model 112 can realize an algorithm that cross-utilizes the data held by multiple local learning models. The federated learning model 112 can be generated by federating local learning models, employing various methods known to those skilled in the art as "federated learning." The term "federated learning" may be synonymous with "integration," for example.

[0020] The local learning model is itself a learning model, and is generated by learning the relationship between multiple customer groups generated from business customer data owned by the business and the consumption behavior corresponding to the business.

[0021] The associative learning model 112 is configured to be able to output potential customer data for the input data received by the input unit 111. That is, for example, when the associative learning model 112 receives a consumption behavior of purchasing a predetermined product as input data, it outputs potential customer data regarding potential customers who are relatively likely to purchase this product.

[0022] The output unit 113 outputs the potential customer data output by the federated learning model 112 to the above-mentioned predetermined external device, etc. The potential customer data includes, for example, an index indicating the possibility that customers in each segment will become customers with the consumption behavior related to the input data according to preset customer segments.

[0023] Next, the processing executed by the data processing device 110 will be described with reference to Fig. 2. Fig. 2 is a flowchart of the data processing method according to the first embodiment. The flowchart shown in Fig. 2 starts, for example, when it is detected that the associative learning model 112 has received input data.

[0024] First, the input unit 111 receives input data relating to a predetermined consumption behavior from an external device or the like communicatively connected to the data processing device 110 (step S11). Upon receiving the input data, the input unit 111 supplies the received input data to the associative learning model 112.

[0025] Next, the associative learning model 112 receives the input data supplied from the input unit 111 (step S12). In other words, the data processing device 110 supplies the input data received by the input unit 111 to the associative learning model 112.

[0026] Next, the associative learning model 112 outputs potential customer data for consumption behavior as an output for the input data. In other words, the data processing device 110 receives the potential customer data as an output from the associative learning model 112 (step S13).

[0027] Next, the output unit 113 outputs the prospective customer data received from the associative learning model 112 to a predetermined output destination (step S14). The output destination is, for example, an external device that accepted the input data. When the output unit 113 outputs the output data, the data processing device 110 ends the series of processes.

[0028] The above has been a description of the data processing device 110. With the above-described configuration, the data processing device 110 can provide a data processing device or the like that contributes to marketing activities simply and suitably.

[0029] Next, a device for generating the associative learning model 112 included in the data processing device 110 will be described with reference to Fig. 3. Fig. 3 is a block diagram of the associative learning model generation device according to the first embodiment.

[0030] The federated learning model generation device 100 may be configured by, for example, a computer or a dedicated device. The federated learning model generation device 100 includes a local learning model acquisition unit 101 and an federated learning model generation unit 102.

[0031] The local learning model acquisition unit 101 acquires a plurality of different local learning models that have learned the relationship between a plurality of customer groups generated from business customer data held by each of a plurality of businesses and the consumption behavior corresponding to the businesses. The local learning model acquisition unit 101 acquires the local learning models from each of the businesses, for example, by communicatively connecting to a computer held by each of the businesses that has the local learning models.

[0032] The federated learning model generation unit 102 federates at least some of the acquired local learning models, thereby generating an federated learning model 112 that receives predetermined consumption behavior of customers as input data and outputs potential customer data for the input data.

[0033] The processing executed by the federated learning model generation device 100 will be described with reference to Fig. 4. Fig. 4 is a flowchart of the federated learning model generation method according to the first embodiment. The flowchart shown in Fig. 4 starts, for example, when the federated learning model generation device 100 detects that it has acquired a local learning model.

[0034] First, the local learning model acquisition unit 101 acquires a plurality of different local learning models that have learned the relationship between a plurality of customer groups generated from business customer data held by a plurality of businesses and the consumption behavior corresponding to the businesses (step S101). The local learning model acquisition unit 101 supplies the acquired local learning models to the federated learning model generation unit 102.

[0035] Next, the federated learning model generation unit 102 generates a federated learning model by federating at least some of the local learning models acquired by the local learning model acquisition unit 101 (step S102). The federated learning model 112 generated by the federated learning model generation unit 102 is configured to accept predetermined consumption behavior of customers as input data and output potential customer data for the input data. When the federated learning model generation unit 102 generates the federated learning model 112, the federated learning model generation device 100 ends the series of processes.

[0036] The above has described the associative learning model generation device 100. According to the above configuration, it is possible to provide an associative learning model and a generation device thereof that contributes to marketing activities simply and suitably.

[0037] This completes the description of embodiment 1. Note that the data processing device 110 and the associative learning model generation device 100 may be separate devices, or may be included in a single device or system.

[0038] The data processing device 110 and the associative learning model generation device 100 each have a processor and a storage device, which are not shown. The storage devices of the data processing device 110 and the associative learning model generation device 100 include storage devices including nonvolatile memory such as flash memory and SSD (Solid State Drive). In this case, the storage device stores a computer program (hereinafter simply referred to as a program) for executing the above-described method. The processor also loads the computer program from the storage device into a buffer memory such as DRAM (Dynamic Random Access Memory) and executes the program.

[0039] Each component of the data processing device 110 and the federated learning model generation device 100 may be realized by dedicated hardware. Furthermore, some or all of the components may be realized by general-purpose or dedicated circuits, processors, etc., or a combination thereof. These may be configured by a single chip, or by multiple chips connected via a bus. Some or all of the components of each device may be realized by a combination of the above-mentioned circuits, etc., and a program. Furthermore, a CPU (Central Processing Unit), GPU (Graphics Processing Unit), FPGA (Field-Programmable Gate Array), etc. may be used as the processor. The description of the configurations described herein may also be applied to other devices or systems described below in this disclosure.

[0040] Furthermore, when some or all of the components of the data processing device 110 and the federated learning model generation device 100 are realized by multiple information processing devices, circuits, etc., the multiple information processing devices, circuits, etc. may be centrally or distributed. For example, the information processing devices, circuits, etc. may be realized as a client-server system, a cloud computing system, or the like, connected via a communication network. Furthermore, the functions of the information processing device 10 may be provided in a SaaS (Software as a Service) format. Furthermore, the above-described method may be stored on a computer-readable medium to cause a computer to execute the above-described method.

[0041] As described above, according to this embodiment, it is possible to provide an associative learning model that contributes to marketing activities simply and suitably, a device for generating the model, and a data processing device that uses the associative learning model.

[0042] <Embodiment 2> Next, an information processing system will be described with reference to Fig. 5. Fig. 5 is a block diagram of a data processing system 1 according to a second embodiment. The data processing system 1 mainly includes a data processing device 120 and a plurality of local data processing devices 200.

[0043] 5 is communicably connected to two local data processing devices 200 via a network N1. One of the local data processing devices 200 is owned by an operator A. The other of the local data processing devices 200 is owned by an operator B.

[0044] For example, business entity A is a car dealer. Business entity A uses the local data processing device 200 that business entity A owns for its own business. For example, customers P1 and P2 visit business entity A to purchase a car. Therefore, business entity A obtains personal information of customers P1 and P2.

[0045] For example, business operator B is a financial service provider that handles certain financial services. Business operator B uses the local data processing device 200 that business operator B owns for its own business. For example, customers P2 and P3 visit business operator B regarding service contracts. Therefore, business operator B obtains personal information of customers P2 and P3.

[0046] In the above situation, business operators A and B manage personal information (also referred to as customer information) of their customers in local data processing device 200. The customer information includes personal information such as the customer's name and address, as well as information about the customer's consumption behavior, such as the products purchased. Business operators A and B generate statistical data that does not include personal information from the customer information they each possess, and generate local learning models from this statistical data. Business operators A and B provide the local learning models they have generated to data processing device 120 via network N1.

[0047] When the data processing device 120 receives the local learning models from each of the businesses A and B, it also generates a federated learning model by federating the received local learning models. The data processing device 120 uses the generated federated learning model to perform predetermined data processing. That is, the data processing device 120 receives data on customer consumption behavior as input data. The data processing device 120 then outputs potential customer data for the received input data.

[0048] Data processing device 120 will be further described with reference to Fig. 6. Fig. 6 is a block diagram of data processing device 120 according to embodiment 2. Data processing device 120 mainly includes associative learning model generation device 100, input unit 111, associative learning model 112, output unit 113, and communication unit 114.

[0049] The federated learning model generation device 100 has the same functions and configuration as those described in embodiment 1. That is, the federated learning model generation device 100 mainly includes a local learning model acquisition unit 101 and an federated learning model generation unit 102.

[0050] The local learning model acquisition unit 101 according to this embodiment acquires local learning models from the local data processing devices 200 owned by each of the operators A and B. The local learning model acquisition unit 101 supplies the acquired multiple local learning models to the federated learning model generation unit 102.

[0051] The federated learning model generation unit 102 according to this embodiment generates the federated learning model 112 by federating the local learning models of the operator A and the operator B. More specifically, the federated learning model generation unit 102 generates the federated learning model by federating at least a portion of the features extracted for each of the multiple local learning models.

[0052] The input unit 111, the associative learning model 112, and the output unit 113 have the same functions and configurations as those described in embodiment 1. The input unit 111 receives input data from an event subject A or an operator B, for example.

[0053] The federated learning model 112 is a federated learning model generated by the federated learning model generation device 100. That is, the federated learning model 112 according to this embodiment includes at least some of the features of the local learning model generated by the operator A and at least some of the features of the local learning model generated by the operator B.

[0054] The output unit 113 outputs the prospective customer data output by the federated learning model 112 to a predetermined output destination. The predetermined output destination is the supplier of the input data received by the input unit 111. For example, when input data is received from business A, the output unit 113 outputs the output data to business A. When input data is received from business B, the output unit 113 outputs the output data to business B.

[0055] The communication unit 114 includes an interface for connecting to the network N1. That is, the communication unit 114 receives various data from the local data processing device 200 via the network N1. The communication unit 114 supplies the received various data to each component of the data processing device 120. The communication unit 114 also receives various data from each component of the data processing device 120 and supplies the received data to the local data processing device 200 via the network N1.

[0056] The local data processing device 200 will be described with reference to Fig. 7. Fig. 7 is a block diagram of the local data processing device according to the second embodiment. The local data processing device 200 mainly includes a customer information acquisition unit 201, a business operator customer data generation unit 202, a local learning model generation unit 203, a communication unit 204, an operation reception unit 205, an information presentation unit 206, a local learning model 210, and a storage unit 220.

[0057] The customer information acquisition unit 201 acquires customer information of the business operator. The customer information may be information received via the operation reception unit 205, or may be information received via the communication unit 204. The customer information acquisition unit 201 supplies the received customer information to the business operator customer data generation unit 202.

[0058] The business customer data generation unit 202 receives customer information from the customer information acquisition unit 201 and generates business customer data 221 in a predetermined format from the received customer information. The business customer data 221 is statistical data that does not include specific personal information. The business customer data generation unit 202 stores the generated business customer data 221 in the memory unit 220.

[0059] The local learning model generation unit 203 generates a local learning model 210 from the business customer data 221 generated by the business customer data generation unit 202. At this time, the local learning model generation unit 203 generates the learning model using a part of the business customer data 221 as input data and another part of the business customer data 221 as output data.

[0060] Various techniques known in the field of machine learning can be used to generate the learning model. For example, the local learning model may have a structure known as a decision tree, or may be configured using a predetermined neural network.

[0061] The communication unit 204 includes an interface for connecting the local data processing device 200 to the network N1. That is, the communication unit 204 receives various data from the data processing device 120 via the network N1. The communication unit 204 supplies the received various data to each component of the local data processing device 200. The communication unit 204 also receives various data from each component of the local data processing device 200 and supplies the received data to the data processing device 120 via the network N1.

[0062] The operation reception unit 205 receives predetermined operations performed by an administrator who manages the local data processing device 200. The predetermined operations include information and instructions input through an input device such as a switch, button, keyboard, mouse, touch panel, or remote controller. More specifically, the operation reception unit 205 receives customer information through an input device operated by the administrator and supplies the received customer information to the customer information acquisition unit 201. Alternatively, the operation reception unit 205 receives predetermined input data related to the local learning model from the administrator. Upon receiving the input data, the operation reception unit 205 supplies the input data to the local learning model 210. Alternatively, the operation reception unit 205 receives predetermined input data related to the federated learning model from the administrator. Upon receiving the input data, the operation reception unit 205 supplies the input data to the data processing device 120 via the communication unit 204.

[0063] The information presentation unit 206 is a means for presenting information relating to various processes performed by the local data processing device 200 to an administrator of the local data processing device 200. Specifically, the information presentation unit 206 is a display device including a liquid crystal panel or organic electroluminescence. The information presentation unit 206 may present predetermined information to other devices (computers, smartphones, etc.) via the communication unit 204.

[0064] The local learning model 210 is a learning model generated by the local learning model generation unit 203. The local learning model 210 is a learning model trained using business customer data 221. For example, the local learning model 210 may receive a predetermined customer group as input and output consumption behavior corresponding to this customer group. For example, the local learning model 210 may receive a business's consumption behavior as input and output a customer group corresponding to this consumption behavior.

[0065] The storage unit 220 is a storage device including a nonvolatile memory such as a hard disk drive (HDD), a solid state drive (SSD), or a flash memory. The storage unit 220 stores business operator customer data 221 generated by the business operator customer data generation unit 202. The business operator customer data 221 includes a common area shared by multiple businesses and a unique area for each of the multiple businesses.

[0066] Next, business customer data generation unit 202 will be described with reference to Fig. 8. Fig. 8 is a diagram showing the processing of business customer data generation unit 202. Fig. 8 shows customer information D10 input to business customer data generation unit 202 and customer statistical data D11 output by business customer data generation unit 202.

[0067] The customer information D10 includes the customer's name, address, age, occupation, and other personal information, and is information that is specific to each customer and is linked to the customer. In addition to the above items, the customer information D10 may also include family structure, marital status, hobbies, preferences, behavioral history, product purchase history, etc.

[0068] The customer statistical data D11 is information that has been processed into statistical data by dividing the customer information D10 into predetermined categories or groups. The customer statistical data D11 does not include personal names or information that can identify individuals. For example, the customer statistical data D11 is divided into customer groups by a predetermined range of addresses, a predetermined range of ages, a predetermined classification of occupations, etc. Therefore, for example, customer group #0001 includes multiple customers that fall into the classification of customer group #0001.

[0069] The business customer data generation unit 202 imports customer information D10 and generates customer statistical data D11 from the imported customer information D10. The customer statistical data D11 is categorized by a common definition across multiple businesses. Therefore, among multiple different businesses, "customer group #0001" is data categorized by a common definition. Furthermore, the customer statistical data D11 does not include personal information. Therefore, the data processing system 1 can use the customer statistical data D11 across multiple businesses without providing the customer information managed by each business to others.

[0070] Next, the business customer data will be described with reference to Fig. 9. Fig. 9 is a diagram showing business customer data 221. The business customer data 221 has the customer statistical data D11 described in Fig. 8 as a common area. The common area refers to a data area set by a common definition across multiple different businesses in the data processing system 1. The common area includes customer groups.

[0071] The business customer data 221 includes a unique area in addition to a common area. The unique area contains data specific to the business of the business and is linked to the customer statistical data D11 in the common area. For example, customer group #0001 shown in Figure 9 has a purchase index of 20% for product type 1 and 42% for product type 2. The purchase index is a statistic of the consumption behavior of customers belonging to the customer group.

[0072] The purchase index may be a probability as shown in Figure 9, or a ranking, score, etc., as long as the associative learning model generation device 100 is capable of generating an associative learning model. In other words, the business customer data 221 includes customer groups included in the business customer data and consumption behaviors linked to the customer groups.

[0073] The local learning model generation unit 203 according to this embodiment generates a local learning model 210 using business customer data 221 as shown in Fig. 9. That is, the local learning model generation unit 203 generates a local learning model trained based on business customer data that includes a common area shared by multiple businesses and a unique area that each of the multiple businesses has.

[0074] Alternatively, the local learning model generation unit 203 generates a local learning model trained based on business customer data including customer groups as a common domain. The local learning model generation unit 203 also generates a local learning model having, as a common domain, data in which customers are classified into multiple groups based on predetermined attributes related to the customers as the customer groups.

[0075] Next, the processing of the local learning model 210 will be described with reference to FIG. 10. FIG. 10 is a diagram showing the processing of the local learning model. The local learning model 210 shown in FIG. 10 is a local learning model owned by business operator A. The local learning model 210 receives data indicating that the automobile type is Type 1 and the color is C2 as input data related to customer consumption behavior. Upon receiving the input data, the local learning model 210 outputs a purchasing index for each customer group as output data corresponding to the input data. In this way, upon receiving the consumption behavior of a business operator as input, the local learning model 210 outputs a customer group corresponding to the consumption behavior.

[0076] Note that the processing performed by the local learning model 210 is not limited to the above content. For example, the local learning model 210 may receive a customer group as input and output consumption behavior corresponding to the customer group.

[0077] Next, the process of generating a federated learning model will be described with reference to Fig. 11. Fig. 11 is a sequence diagram showing the process executed by the federated learning model generation system. Fig. 11 shows a configuration for generating a federated learning model, in which local data processing device 200 owned by business operator A, local data processing device 200 owned by business operator B, and federated learning model generation device 100 are shown as federated learning model generation system 2.

[0078] That is, the federated learning model generation system 2 includes a plurality of local data processing devices 200 and a federated learning model generation device 100. The local data processing device 200 generates a local learning model 210 for each business operator from business operator customer data managed by each business operator. The federated learning model generation device 100 acquires the local learning models 210 from each of the plurality of businesses and generates a federated learning model 112.

[0079] In the sequence diagram shown in FIG. 11, first, the local data processing device 200 of the business A acquires customer information (step S201). Next, the local data processing device 200 of the business A generates business customer data from the acquired customer information (step S202). Furthermore, the local data processing device 200 of the business A generates a local learning model from the business customer data (step S203).

[0080] Similarly to the case of business operator A, the local data processing device 200 of business operator B also first acquires customer information (step S211) and then generates business operator customer data (step S212).Then, the local data processing device 200 of business operator B generates a local learning model from the generated business operator customer data (step S213).

[0081] Next, the federated learning model generation device 100 acquires the local learning model of the operator A and the local learning model of the operator B (step S221). Then, the federated learning model generation device 100 generates a federated learning model by federating the acquired local learning models (step S222).

[0082] The federated learning model generation system 2 has been described above with reference to Fig. 11. As described above, the federated learning model generation device 100 cooperates with the local data processing device 200 to configure the federated learning model generation system 2.

[0083] Next, the configuration of the federated learning model will be described. FIG. 12 is a diagram showing the configuration of the federated learning model. The federated learning model 112 shown in FIG. 12 is an example configured using a decision tree. When federated learning model 112 shown in FIG. 12 receives input data, it processes the input data using the multiple nodes that it configures. Then, federated learning model 112 outputs output data shown at the bottom of FIG. 12. Note that the diagram shown in FIG. 12 is a schematic diagram of the data structure of a decision tree, and the number of nodes, the number of layers, etc. are not limited to the configuration of FIG. 12.

[0084] Federated learning model 112 has element E1 and element E2. Element E1 includes at least some of the features of local learning model 210 owned by operator A. Element E2 includes at least some of the features of local learning model 210 owned by operator B. In this way, federated learning model 112 is generated by extracting at least some of the features owned by each local learning model and associating the extracted features. In other words, federated learning model generation unit 102 of federated learning model generation device 100 generates federated learning model 112 by associating at least some of the features extracted for each of multiple local learning models.

[0085] 12 shows an example of a decision tree, the associative learning model 112 of the data processing system 1 is not limited to a decision tree. The associative learning model 112 may be configured as a predetermined perceptron or neural network.

[0086] Next, the processing executed by federated learning model 112 will be described with reference to Fig. 13. Fig. 13 is a diagram showing the processing of the federated learning model. Federated learning model 112 shown in Fig. 13 is generated by federating a local learning model owned by business operator A and a local learning model owned by business operator B. Therefore, federated learning model 112 can accept data on consumer behavior related to business operator A's business as input data, and can also accept data on consumer behavior related to business operator B's business.

[0087] In the upper part of Fig. 13, federated learning model 112 receives input data related to automobile consumption behavior related to the business of business operator A. In this case, federated learning model 112 outputs a purchasing index for each customer group as output data corresponding to the received automobile type and specifications.

[0088] In the lower part of Figure 13, federated learning model 112 receives input data related to consumption behavior of financial services related to the business of business operator B. In this case, federated learning model 112 outputs a purchasing index for each customer group as output data corresponding to the specifications of the received financial services.

[0089] In this way, the federated learning model 112 accepts input data related to the consumption behavior of each of multiple businesses and outputs prospective customer data corresponding to the accepted input data. The federated learning model 112 is configured to be able to output prospective customer data for the consumption behavior when it accepts input data related to a specific consumption behavior. At this time, the business customer data held by business A, the business customer data held by business B, and the prospective customer data output by the federated learning model 112 each include a common customer group. The federated learning model 112 outputs prospective customer data including at least a portion of the multiple customer groups. At this time, the federated learning model 112 may output prospective customer data including an estimated purchasing index that indicates the tendency of the consumption behavior of the customer group.

[0090] Next, the processing executed by data processing system 1 will be described with reference to Fig. 14. Fig. 14 is a sequence diagram showing the processing executed by data processing system 1. The sequence diagram shown in Fig. 14 shows a flow in which local data processing device 200 owned by business operator B, which is a financial service business operator, receives input data for federated learning model 112, and in response, federated learning model 112 outputs output data serving as an answer to local data processing device 200.

[0091] First, the local data processing device 200 receives a predetermined query from a user who is a user of the data processing system 1 (step S301). For example, the query may be "Who are the potential customers for automobile type 1?"

[0092] Next, the local data processing device 200 generates input data to be transmitted to the data processing device 120 from the content of the query, and transmits the generated input data to the data processing device 120 (step S302).

[0093] Next, the data processing device 120 acquires input data from the local data processing device 200 (step S11), and supplies the acquired input data to the associative learning model 112 (step S12).

[0094] When federated learning model 112 receives input data, it outputs potential customer data, which is output data corresponding to the received input data. Data processing device 120 receives the output data from federated learning model 112 (step S13) and transmits the received output data to local data processing device 200 (step S14).

[0095] Next, the local data processing device 200 obtains the output data from the data processing device 120 (step S303), and uses the obtained output data to present an answer to the user (step S304). The answer includes the customer group.

[0096] The above describes the processing executed by the data processing system 1. In the data processing system 1, the federated learning model 112 receives input related to consumer behavior and outputs potential customers. The federated learning model does not use personal information, but handles information on customer groups grouped according to predetermined criteria. Therefore, the data processing system 1 can link data in the unique areas of different businesses via data in the common area through federated learning. Therefore, the data processing system 1 can output potential customer data to each business using data in the common area spanning multiple businesses. In other words, the data processing system 1 can estimate potential customer data related to the business of each business with high accuracy.

[0097] 14, the data processing system 1 can send a query to the data processing device 120 using consumption behavior related to a business other than its own business as input data, and obtain potential customer data that serves as a response to the query. This allows a business using the data processing system 1 to obtain potential customer data on consumption behavior that is difficult to obtain from its own business. Therefore, the data processing system 1 can contribute to expanding the scope of the business's marketing activities.

[0098] The administrator who manages data processing device 120 may be business operator A or business operator B, or may be a person different from business operator A and business operator B. Federated learning model 112 may be owned by the administrator of data processing device 120.

[0099] As described above, according to the second embodiment, it is possible to provide an associative learning model that contributes to marketing activities simply and suitably, a device for generating the model, and a data processing device that uses the associative learning model.

[0100] <Embodiment 3> Next, a description will be given of embodiment 3. Embodiment 3 differs from the above-described embodiments in that it has a function of updating the local learning model and the federated learning model.

[0101] 15 is a block diagram of a data processing device 130 according to the third embodiment. The data processing device 130 shown in FIG.

[0102] The update control unit 115 controls acquisition of a local learning model when the local learning model is updated. Specifically, the update control unit 115 appropriately acquires update information from multiple local data processing devices. Then, when detecting that the local learning model has been updated, the update control unit 115 instructs the local learning model acquisition unit 101 to acquire update data from the local data processing device having the updated local learning model.

[0103] Furthermore, the update control unit 115 controls updating of the federated learning model using the updated local learning model. That is, the update control unit 115 instructs the federated learning model generation unit 102 to update the federated learning model using the acquired update data.

[0104] 16 is a block diagram of a local data processing device 300 according to the third embodiment. The local data processing device 300 differs from the above-described local data processing device 200 in that it includes an update information management unit 207. The local data processing device 300 also stores update information 222 in a storage unit 220.

[0105] The local learning model 210 is updated, for example, when the customer information acquisition unit 201 acquires new customer data. Alternatively, the local learning model 210 may be updated in conjunction with an update of the business customer data 221.

[0106] When the local learning model 210 is updated, the update information management unit 207 updates the update information 222 stored in the storage unit 220 to manage the version information or update date and time of the local learning model 210. The update information management unit 207 may have a function of notifying the data processing device 130 that the local learning model 210 has been updated.

[0107] The update information 222 stored in the storage unit 220 may include version information of the local learning model 210, or may include the latest update date and time of the local learning model 210.

[0108] Next, the update process executed by the data processing system 1 will be described with reference to Fig. 17. Fig. 17 is a sequence diagram of the update method according to the third embodiment.

[0109] First, the update information management unit 207 of the local data processing device 300 determines whether the customer information or business customer data 221 has been updated (step S401). If it is determined that the customer information or business customer data 221 has not been updated (step S401: NO), the local data processing device 300 repeats step S401. If it is determined that the customer information or business customer data 221 has been updated (step S401: YES), the local data processing device 300 updates the local learning model 210 using the update data (step S402).

[0110] Next, the data processing device 130 obtains the updated local learning model 210 from the local data processing device 300 (step S403).

[0111] Next, the data processing device 130 updates the federated learning model 112 using the updated local learning model 210 (step S404).

[0112] The above describes the process in which data processing device 130 updates the federated learning model using an updated local learning model. Data processing system 1 according to this embodiment can contribute favorably to marketing activities by appropriately updating federated learning model 112. Therefore, this embodiment can provide an federated learning model that contributes simply and favorably to marketing activities, a device for generating the federated learning model, and a data processing device that uses this federated learning model.

[0113] <Example of hardware configuration> Hereinafter, a case will be described in which each functional configuration of the determination device according to the present disclosure is realized by a combination of hardware and software.

[0114] FIG. 18 is a block diagram illustrating an example of a hardware configuration of a computer. The management device of the present disclosure can realize the above-described functions by a computer 500 including the hardware configuration shown in the figure. The computer 500 may be a portable computer such as a smartphone or tablet terminal, or a stationary computer such as a PC. The computer 500 may be a dedicated computer designed to realize each device, or may be a general-purpose computer. The computer 500 can realize desired functions by installing a predetermined program.

[0115] The computer 500 has a bus 502, a processor 504, a memory 506, a storage device 508, an input / output interface 510 (an interface is also called an I / F (Interface)), and a network interface 512. The bus 502 is a data transmission path through which the processor 504, the memory 506, the storage device 508, the input / output interface 510, and the network interface 512 transmit and receive data to and from each other. However, the method of connecting the processor 504 and other components to each other is not limited to a bus connection.

[0116] The processor 504 is one of various processors such as a CPU, a GPU, an FPGA, etc. The memory 506 is a main storage device realized using a RAM (Random Access Memory) or the like.

[0117] The storage device 508 is an auxiliary storage device realized using a hard disk, an SSD, a memory card, a ROM (Read Only Memory), or the like. The storage device 508 stores programs for realizing desired functions. The processor 504 reads the programs into the memory 506 and executes them to realize the respective functional components of each device.

[0118] The input / output interface 510 is an interface for connecting the computer 500 with input / output devices. For example, the input / output interface 510 is connected to an input device such as a keyboard and an output device such as a display device.

[0119] The network interface 512 is an interface for connecting the computer 500 to a network.

[0120] Although an example of a hardware configuration in the present disclosure has been described above, the above-described embodiment is not limited to this. Any processing in the present disclosure can also be realized by causing a processor to execute a computer program.

[0121] In the above examples, the program includes instructions (or software code) that, when loaded into a computer, cause the computer to perform one or more functions described in the embodiments. The program may be stored on a non-transitory computer-readable medium or a tangible storage medium. By way of example and not limitation, computer-readable medium or tangible storage medium includes random-access memory (RAM), read-only memory (ROM), flash memory, solid-state drive (SSD) or other memory technology, CD-ROM, digital versatile disc (DVD), Blu-ray disc or other optical disk storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage device. The program may also be transmitted on a transitory computer-readable medium or communication medium. By way of example and not limitation, transitory computer-readable medium or communication medium includes electrical, optical, acoustic, or other forms of propagated signals.

[0122] Although the present invention has been described above with reference to the embodiments, the present invention is not limited to the above. Various modifications that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the invention.

[0123] Some or all of the above embodiments can be described as, but are not limited to, the following supplementary notes. (Appendix 1) a local learning model acquisition unit that acquires a plurality of different local learning models that have learned the relationship between a plurality of customer groups generated from business customer data held by a plurality of businesses and consumption behaviors corresponding to the businesses; and an federated learning model generation unit that generates an federated learning model by federating at least some of the acquired local learning models, the federated learning model receiving the predetermined consumption behavior of customers as input data and outputting potential customer data for the input data. Federated learning model generator. (Appendix 2) the federated learning model generation unit generates the federated learning model that outputs the prospective customer data including at least a portion of the plurality of customer groups. 2. The apparatus for generating a federated learning model according to claim 1. (Appendix 3) the federated learning model generation unit generates the federated learning model that outputs the potential customer data including an estimated purchase index that indicates the tendency of the consumption behavior of the customer group. 3. The apparatus for generating a federated learning model according to claim 2. (Appendix 4) the local learning model acquisition unit, upon receiving the customer group as an input, acquires the local learning model that outputs the consumption behavior corresponding to the customer group; 4. The federated learning model generation device according to claim 3. (Appendix 5) the local learning model acquisition unit, upon receiving the consumption behavior of the business operator as an input, acquires the local learning model that outputs the customer group corresponding to the consumption behavior; 4. The federated learning model generation device according to claim 3. (Appendix 6) the federated learning model generation unit generates the federated learning model by federating at least some of the features extracted for each of the plurality of local learning models. 2. The apparatus for generating a federated learning model according to claim 1. (Appendix 7) The local learning model acquisition unit acquiring the local learning model trained based on the business customer data, each of which includes a common area shared by the plurality of businesses and a unique area shared by each of the plurality of businesses; 7. The associative learning model generation device according to any one of Supplementary notes 1 to 6. (Appendix 8) The local learning model acquisition unit obtaining the local learning model trained based on the business customer data including the customer group as the common domain; 8. The federated learning model generation device according to claim 7. (Appendix 9) The local learning model acquisition unit acquiring the local learning model having, as the customer group, data obtained by classifying the customers into a plurality of groups based on predetermined attributes of the customers as the common domain; 9. The federated learning model generation device according to claim 8. (Appendix 10) a local data processing device that generates the local learning model for each of the businesses based on the business customer data managed by each of the businesses; and a federated learning model generation device according to claim 7, which acquires the local learning models from the plurality of businesses and generates the federated learning model. Federated learning model generation system. (Appendix 11) an update control unit that acquires the local learning model when the local learning model is updated and controls updating the federated learning model using the updated local learning model; 11. The federated learning model generation system of claim 10. (Appendix 12) The computer Obtaining a plurality of different local learning models that have learned the relationship between a plurality of customer groups generated from business customer data owned by a plurality of businesses and consumption behaviors corresponding to the businesses; by associating at least some of the acquired local learning models, a federated learning model is generated that receives the predetermined consumption behavior of the customer as input data and outputs potential customer data for the input data. Federated learning model generation method. (Appendix 13) Obtaining a plurality of different local learning models that have learned the relationship between a plurality of customer groups generated from business customer data owned by a plurality of businesses and consumption behaviors corresponding to the businesses; by associating at least some of the acquired local learning models, a federated learning model is generated that receives the predetermined consumption behavior of the customer as input data and outputs potential customer data for the input data. Executing the method for generating an associative learning model on a computer; program. (Appendix 14) A federated learned model generated by federating a plurality of different locally trained models, The local learning model includes at least some of the characteristics of each of a plurality of different businesses that have learned the relationship between business customer data and predetermined consumption behavior based on the business customer data managed by each of the plurality of different businesses, When input data relating to a predetermined consumption behavior is received, potential customer data for the consumption behavior can be output; The business customer data and the potential customer data each include a common customer group. Associative learning model.

[0124] This application claims priority based on Japanese Patent Application No. 2022-089722, filed on June 1, 2022, the disclosure of which is incorporated herein by reference in its entirety. [Industrial Applicability]

[0125] The present disclosure can be used, for example, as a data processing device for businesses to carry out marketing activities. [Explanation of symbols]

[0126] 1. Data Processing System 2. Federated Learning Model Generation System 100 Associative learning model generation device 101 Local learning model acquisition unit 102 Associative Learning Model Generation Unit 110 Data processing device 111 Input section 112 Associative Learning Model 113 Output section 114 Communications Department 115 Update control section 120 Data Processing Device 130 Data Processing Device 200 Local Data Processing Device 201 Customer Information Acquisition Department 202 Business Customer Data Generation Department 203 Local Learning Model Generation Unit 204 Communications Department 205 Operation reception section 206 Information Presentation Department 207 Update information management department 210 Local Learning Model 220 Storage section 221 Business Customer Data 222 Update information 300 Local Data Processing Device 500 computers 504 processor 506 memory 508 Storage Devices 510 Input / Output Interface 512 network interface A10 readable area G10 Guidance Image G11 First guide image G12 2nd lead image N1 Network P1 Customer P2 Customer D10 Customer Information D11 Customer Statistical Data

Claims

1. a local learning model acquisition means for acquiring a plurality of different local learning models that have learned the relationship between a plurality of customer groups generated from business customer data held by a plurality of businesses and consumption behaviors corresponding to the businesses; and an association learning model generation means for generating an association learning model that receives the predetermined consumption behavior of customers as input data and outputs potential customer data for the input data by associating at least some of the acquired local learning models. Federated learning model generator.

2. the federated learning model generation means generates the federated learning model that outputs the prospective customer data including at least a portion of the plurality of customer groups. The apparatus for generating a federated learning model according to claim 1 .

3. the federated learning model generation means generates the federated learning model that outputs the potential customer data including an estimated purchase index that indicates the tendency of the consumption behavior of the customer group. The apparatus for generating an associative learning model according to claim 2 .

4. 4. The federated learning model generation device according to claim 3, wherein the local learning model acquisition means, upon receiving the customer group as an input, acquires the local learning model that outputs the consumption behavior corresponding to the customer group.

5. the local learning model acquisition means, when receiving the consumption behavior of the business operator as an input, acquires the local learning model that outputs the customer group corresponding to the consumption behavior; The associative learning model generating device according to claim 3 .

6. the federated learning model generation means generates the federated learning model by federating at least some of the features extracted for each of the plurality of local learning models. The apparatus for generating an associative learning model according to any one of claims 1 to 5.

7. a local data processing device that generates the local learning model for each of the businesses based on the business customer data managed by each of the businesses; A federated learning model generation system comprising: a federated learning model generation device according to any one of claims 1 to 5, which acquires the local learning models from each of the plurality of businesses and generates the federated learning model.

8. The computer Obtaining a plurality of different local learning models that have learned the relationship between a plurality of customer groups generated from business customer data owned by a plurality of businesses and consumption behaviors corresponding to the businesses; by associating at least some of the acquired local learning models, a federated learning model is generated that receives the predetermined consumption behavior of the customer as input data and outputs potential customer data for the input data. Federated learning model generation method.

9. Obtaining a plurality of different local learning models that have learned the relationship between a plurality of customer groups generated from business customer data owned by a plurality of businesses and consumption behaviors corresponding to the businesses; by associating at least some of the acquired local learning models, a federated learning model is generated that receives the predetermined consumption behavior of the customer as input data and outputs potential customer data for the input data. Executing the method for generating an associative learning model on a computer; program.

10. A federated learned model generated by federating a plurality of different locally trained models, The local learning model includes at least some of the characteristics of each of a plurality of different businesses that have learned the relationship between the business customer data and a predetermined consumption behavior based on the business customer data managed by each of the plurality of different businesses, When input data relating to the predetermined consumption behavior is received, the computer is caused to function to output potential customer data for the consumption behavior; The business customer data and the potential customer data each include a common customer group. Associative learning model.

Images