Integration systems, organizational servers, integration methods, analysis methods, and programs

Abstract

This integration system comprises an acquisition unit, a integration unit, and a provision unit. The acquisition unit acquires a concealed parameter of a local model that is trained for each of a plurality of organization servers on the basis of consumer purchase history and consumer attribute information of each of the plurality of organization servers, the local model being capable of outputting, upon input of attribute information, information relating to the tendency of consumption of a consumer corresponding to the attribute represented by the inputted attribute information. The integration unit integrates the concealed parameters concerning each of the plurality of organization servers, thereby generating a global model. The provision unit provides the integrated parameters of the generated global model to the plurality of organization servers.

Description

Technical Field

[0001] This disclosure relates to an integrated system and the like.

Background Art

[0002] In some cases, a consumer's consumption trend may be analyzed using the consumer's purchase history.

[0003] Also, Patent Document 1 describes integrating a learning model by secure computation.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] For example, it may be difficult to analyze a consumer's consumption trend based only on the purchase history of the consumers owned by the company itself. Also, for example, there are organizations that do not have the purchase history of consumers. However, since a consumer's purchase history corresponds to corporate secrets and personal information, it may be difficult for each organization to provide purchase history and the like to other organizations.

[0006] An example of the object of this disclosure is to provide an integrated system and the like that can improve the accuracy of analyzing consumption trends.

Means for Solving the Problems

[0007] An integrated system in one aspect of this disclosure includes: an acquisition means for each of a plurality of organizational servers, which is a local model learned based on the consumer purchase history and attribute information of the consumer, and which, when attribute information is input, can output information regarding the consumption trends of a consumer corresponding to the attribute represented by the input attribute information; an integration means for generating a global model by integrating the confidential parameters for each of the plurality of organizational servers; and a provision means for providing the plurality of organizational servers with the integrated parameters of the generated global model.

[0008] An organizational server in one aspect of this disclosure includes: a model generation means that learns a local model capable of outputting information on the consumer's consumption trends corresponding to the attributes represented by the input attribute information, based on the consumer's purchase history and attribute information of the consumer; anonymization means that conceals the parameters of the local model using secure computation; output means that output the concealed parameters of the local model to an integrated system; parameter acquisition means that acquires parameters of a global model integrated in the integrated system from the integrated system; and analysis means that inputs attribute information into the global model based on the acquired parameters and acquires information on the consumer's consumption trends corresponding to the attributes represented by the input attribute information from the global model.

[0009] An integration method in one aspect of this disclosure involves obtaining anonymized parameters of a local model, which is learned for each of a plurality of organizational servers based on the consumer purchase history and attribute information of the consumer, and which, when attribute information is input, can output information regarding the consumer's consumption trends corresponding to the attribute represented by the input attribute information; integrating the anonymized parameters for each of the plurality of organizational servers to generate a global model; and providing the plurality of organizational servers with the integrated parameters of the generated global model.

[0010] An analysis method in one aspect of this disclosure involves learning a local model that, when attribute information is input, outputs information on the consumer's consumption trends corresponding to the attribute represented by the input attribute information, based on the consumer's purchase history and attribute information of the consumer; concealing the parameters of the local model using secure computation; outputting the concealed parameters of the local model to an integrated system; obtaining the parameters of a global model integrated in the integrated system from the integrated system; inputting attribute information into the global model based on the obtained parameters; and obtaining information on the consumer's consumption trends corresponding to the attribute represented by the input attribute information from the global model.

[0011] A program in one aspect of this disclosure causes a computer to perform a process that involves obtaining confidential parameters of a local model, which is learned for each of a plurality of organizational servers based on the consumer purchase history and attribute information of said consumers, and which, when attribute information is input, can output information regarding the consumption trends of consumers corresponding to the attributes represented by the input attribute information; integrating the confidential parameters for each of the plurality of organizational servers to generate a global model; and providing the plurality of organizational servers with the integrated parameters of the generated global model.

[0012] A program in one aspect of this disclosure learns a local model on a computer that, when attribute information is input, outputs information about the consumer's consumption trends corresponding to the attribute represented by the input attribute information, based on the consumer's purchase history and attribute information of the consumer. The parameters of the local model are made secret using secure computation, the secreted parameters of the local model are output to the integrated system, the parameters of the global model integrated in the integrated system are obtained from the integrated system, attribute information is input to the global model based on the obtained parameters, and information regarding consumer consumption trends corresponding to the attributes represented by the input attribute information is obtained from the global model.

[0013] Each program may be stored on a non-temporary storage medium that is readable by the computer. [Effects of the Invention]

[0014] According to this disclosure, it is possible to improve the accuracy of consumer trend analysis. [Brief explanation of the drawing]

[0015] [Figure 1] This is a block diagram showing an example configuration of an information processing system according to Embodiment 1. [Figure 2] This is a flowchart showing an example of operation of the integrated system according to Embodiment 1. [Figure 3] This is an explanatory diagram showing an example of how to connect each device in an information processing system. [Figure 4] This is a block diagram showing an example configuration of an information processing system according to Embodiment 2. [Figure 5] This is an explanatory diagram showing an example of displaying analysis results on a terminal device. [Figure 6] This is an explanatory diagram showing an example of displaying recommended product information on a terminal device. [Figure 7] This is a flowchart showing an example of an operation of the information processing system according to Embodiment 2. [Figure 8] This is an explanatory diagram showing an example of a computer hardware configuration. [Modes for carrying out the invention]

[0016] Referring to the drawings below, embodiments of an integrated system, an organization server, an integration method, an analysis method, a program, and a non-transitory recording medium for recording the program according to the present disclosure will be described in detail. These embodiments do not limit the disclosed technology.

[0017] (Embodiment 1) First, in Embodiment 1, the basic functions of the information processing system will be described. FIG. 1 is a block diagram showing a configuration example of the information processing system according to Embodiment 1. The information processing system 1 is a system for integrating parameters of a local model that outputs information regarding the consumption trends of consumers corresponding to the attributes represented by the input attribute information when the attribute information of consumers is input. The information processing system 1 includes an integrated system 10 and a plurality of organization servers 11. In the following description, when the local model or the global model is not specified, it may be simply referred to as a model.

[0018] The attribute information of consumers represents the attributes of consumers for each consumer. Examples of consumer attributes include age, generation, family composition, occupation, workplace, gender, income, residential area, residence, hobbies, and the like.

[0019] The purchase history includes information on the products purchased by consumers for each consumer. More specifically, the purchase history may store, for each consumer, information on products, purchase quantities, purchase dates, etc. in association with each other. For example, information on products of apparel companies, information on products of supermarkets, and information on products of real estate companies are expected to be different from each other. Thus, the product information may vary depending on characteristics of the target company, etc. Note that services may be used instead of products.

[0020] The attribute information and the purchase history may be associated with each other by consumer identification information that identifies consumers.

[0021] The model is designed to analyze customer consumption trends. For example, the model takes consumer attribute information as input and outputs information related to customer consumption trends. This information could include information about products that match the consumer's attribute information, or information about the characteristics of products that match the consumer's attribute information. Specifically, the information about customer consumption trends could be information about specific recommended products, or information about product characteristics. For example, in the case of food products, product characteristics could be things like strong flavor or mild flavor.

[0022] The models include, but are not limited to, decision tree models, linear regression models, logistic regression models, and neural network models.

[0023] Furthermore, the model described here may be a horizontally associative learning model or a vertically associative learning model, and is not particularly limited. Horizontally associative learning is a learning process that does not identify individual consumers. In horizontally associative learning, each organization holds the same dataset of purchase history, although the purchase history held by each organization may be for different consumers. In horizontally associative learning, the performance of the model can be improved by increasing the overall amount of training data. On the other hand, vertically associative learning is a learning process that identifies individual consumers. In vertically associative learning, each organization holds the same or different datasets of purchase history for the same consumer.

[0024] Here, an organization refers to, for example, a retail company or a department within a company. For example, an organization is a customer of the providing company. If an organization server 11 is not specified, it is simply referred to as an organization server 11. An organization server 11 is a device that learns a local model using consumer purchase history and consumer attribute information held by an organization, and analyzes consumer trends using a global model. Organization servers 11 can be separated by organization, for example, and in Figure 1, the number of organization servers 11 is shown as two as an example, but is not particularly limited.

[0025] Here, parameters are values ​​acquired after the model has been trained. If the purchase history includes product information where the attribute represented by the attribute information is the primary customer, then the weights for this product information are the parameters. More specifically, for example, if the attribute information is age, and the purchase history includes product information where this age group is the primary customer, then the weights for this product information are the parameters.

[0026] The integrated system 10 includes an acquisition unit 101, an integration unit 102, and a provision unit 103. The organization server 11 includes a local model generation unit 111, an anonymization unit 112, an output unit 113, a parameter acquisition unit 114, and an analysis unit 116. The organization server 11 also includes a model storage unit 115. The model storage unit 115 stores each model, for example.

[0027] First, let's briefly explain the organization server 11. The local model generation unit 111 generates a local model based on customer attribute information and customer purchase history. For example, the local model generation unit 111 stores the generated local model in the model storage unit 115 or the like.

[0028] Next, the concealment unit 112 conceals the parameters of the local model generated by the local model generation unit 111 using secure computation. Concealment is synonymous with encryption. Secure computation is a technique for performing calculations while concealing the data. The output unit 113 outputs the concealed parameters to the integrated system 10. The integrated system 10, which is the basic function of Embodiment 1, will be described in detail below.

[0029] In the integrated system 10, the acquisition unit 101 acquires the concealed parameters of the local model from each of the multiple organization servers 11. Specifically, for example, the acquisition unit 101 receives the concealed parameters of the trained local model from each of the multiple organization servers 11. The acquisition unit 101 may receive the concealed parameters triggered by an operation by the service provider to integrate the model parameters. Alternatively, the acquisition unit 101 may periodically receive the concealed parameters from the organization servers 11.

[0030] In the integrated system 10, the integration unit 102 generates a global model by integrating multiple concealed parameters. The integration unit 102 may change the weights of the parameters corresponding to each local model depending on the characteristics of each local model when integrating the multiple concealed parameters.

[0031] The integration of multiple concealed parameters involves the information processing system 1 performing machine learning (federated learning) in a distributed state across each organizational server 11, and then integrating the parameters of the trained local models. Here, integration refers to a method of combining data from different entities so that it can be widely utilized. In this way, the information processing system 1 can integrate parameters while each parameter remains concealed. As for secure computation methods, special encryption corresponding to specific processing, such as homomorphic encryption, a trusted execution environment that processes in an isolated state on hardware, or multi-party computation that performs computation while secretly sharing the data across multiple servers (secret sharing computation) can be used.

[0032] An example of using secure computation in multi-party computing is as follows: When using secure computation in multi-party computing, the integrated system 10 may have multiple servers to realize the functions of the integrated unit 102. For example, the number of servers may be three. For example, one of the servers secretly distributes confidential data a, which is a parameter obtained from an arbitrary organization server 11, into distributed values ​​x1, y1, ... and sends x1, y1, ... to different servers. Also, one of the servers secretly distributes confidential data b, which is a parameter obtained from another organization server 11, into distributed values ​​x2, y2, ... and sends x2, y2, ... to different servers. The multiple servers communicate with each other and proceed with the computation while confidential data a and confidential data b remain secretly distributed. Finally, the distributed values ​​u, v, ... of the output, which are the computation results of each server, are collected and restored to obtain the computation result F(a, b). This computation result becomes the parameter of the global model, which integrates the parameters of each local model. Multi-party computation eliminates the need for cryptographic key management and isolated environments, resulting in faster computation.

[0033] Here, for example, the integration unit 102 may integrate all the concealed parameters obtained from each of the multiple organization servers 11. Alternatively, for example, the integration unit 102 may select some of the concealed parameters from the multiple organization servers 11 and integrate the selected parameters. When selecting some parameters, the integration unit 102 may select some of the concealed parameters from the concealed parameters for each of the multiple organization servers 11 based on the organizational characteristics of the organization servers 11. Organizational characteristics include, for example, the industry of the organization, the type of business the organization operates in, and the types of products sold. Specifically, for example, the integration unit 102 may select some of the concealed parameters from the concealed parameters for each of the multiple organization servers 11 whose organizational characteristics are specific characteristics. The specific characteristics are not particularly limited, but for example, they may be the characteristics of the requesting organization that requested the global model. Then, the integration unit 102 integrates the selected some of the concealed parameters. This makes it possible to obtain global model parameters for multiple organizations with similar organizational characteristics.

[0034] The provisioning unit 103 provides, for example, the integrated parameters of the generated global model to each organization server 11. Specifically, as a method of provision, for example, the provisioning unit 103 sends the parameters of the global model to each organization server 11. The provisioning unit 103 may also send the parameters of the global model to organization servers 11 that did not provide the confidential parameters of the local model to the integration system 10.

[0035] Furthermore, the provisioning unit 103 may, for example, send the parameters of the global model to the requesting organization server 11 when a request for the global model is received from the organization server 11. Also, when selecting some confidential parameters, the provisioning unit 103 may provide the integrated parameters to the organization server 11 that is the source of some of the confidential parameters among the multiple organization servers 11. Also, for example, when selecting some confidential parameters based on the characteristics of a company, the provisioning unit 103 may provide the integrated parameters to the organization server 11 that matches the characteristics of the company.

[0036] Then, in the organization server 11, the parameter acquisition unit 114 acquires the parameters of the global model. Specifically, as a method of acquisition, the parameter acquisition unit 114 receives the parameters of the global model from the integrated system 10, for example.

[0037] The timing at which the parameter acquisition unit 114 acquires parameters is not particularly limited. For example, the parameter acquisition unit 114 may receive confidential parameters that are periodically transmitted from the integrated system 10. The parameter acquisition unit 114 may request the integrated system 10 to transmit parameters and, in response to the request, receive global model parameters from the integrated system 10.

[0038] The parameter acquisition unit 114 can obtain the global model by setting the acquired parameters to the parameters of the local model. This allows each organization to use the global model. The parameter acquisition unit 114 may also acquire the global model. For example, the parameter acquisition unit 114 stores the global model in the model storage unit 115 or the like.

[0039] The analysis unit 116 then inputs consumer attribute information into the global model and obtains information on consumer consumption trends from the global model. Examples of how each organization uses the global model will be explained in detail in Embodiment 2.

[0040] (flowchart) Figure 2 is a flowchart illustrating an example of the operation of the integrated system 10 according to Embodiment 1. In the integrated system 10, the acquisition unit 101 acquires confidential parameters for each of the multiple local models (step S101). The integration unit 102 integrates the confidential parameters (step S102). The provision unit 103 provides the parameters for the global model (step S103). The integrated system 10 terminates processing.

[0041] As mentioned above, referring to purchase history and consumer attribute information held by each organization may involve confidential or personal information of the organization. For this reason, as previously stated, it is difficult for each organization to provide purchase history or consumer attribute information to other organizations. Thus, when analyzing consumer spending trends using the purchase history and attribute information of a single organization, there is a problem in that the accuracy of the analysis may be low.

[0042] As described above, in Embodiment 1, the integrated system 10 obtains the concealed parameters of a local model learned from each of the multiple organization servers 11 based on the consumer's purchase history and consumer attribute information, and which can output information on consumer consumption trends when consumer attribute information is input. The integrated system 10 then integrates the concealed parameters and provides the integrated parameters to the multiple organization servers 11. For example, the information on consumption trends is information on products that match the consumer's attribute information. The product information may be, for example, the product name or product characteristics. This allows the integrated system 10 to obtain a global model without acquiring operational data from each organization. Furthermore, each organization can perform analysis using the global model generated using the purchase history and consumer attribute information of various organizations without having to provide purchase history or consumer attribute information. Therefore, the accuracy of consumer trend analysis can be improved.

[0043] Furthermore, the integration system 10 selects some parameters from the anonymized parameters of multiple organization servers 11 based on the characteristics of each organization, and integrates the selected parameters. For example, in the process of selecting some parameters, the integration system 10 may select parameters that have the same or similar characteristics as the organizations. Therefore, the accuracy of the analysis of consumption trends can be improved.

[0044] (Embodiment 2) Next, Embodiment 2 will be described in detail with reference to the drawings. Embodiment 2 describes an example of analyzing consumer trends. In Embodiment 2, explanations that overlap with the above description will be omitted to the extent that the description of Embodiment 2 does not become unclear.

[0045] Figure 3 is an explanatory diagram showing an example of the connection of each device in the information processing system 1. The information processing system 1 comprises an integrated system 10, a plurality of organizational servers 11, and a plurality of terminal devices 12. The integrated system 10, the plurality of organizational servers 11, and the plurality of terminal devices 12 are connected, for example, via a communication network NT. Note that the communication network NT may consist of multiple communication networks. For example, the communication network to which the integrated system 10 and organizational server 11a are connected may be different from the communication network to which the integrated system 10 and organizational server 11b are connected.

[0046] If the terminal device 12 is not specified, it is simply referred to as terminal device 12. For example, terminal device 12 is a device that receives notifications of consumer trends. Terminal device 12 may be, for example, the terminal device 12 of an organization's representative. Terminal device 12 may be, for example, separate for each organization. Alternatively, terminal device 12 may be, for example, the terminal device 12 of a consumer who is a customer of the organization. Terminal device 12 may be, for example, separate for each consumer. The type of terminal device 12 is not particularly limited and may include smartphones, tablet terminal devices 12, PCs (Personal Computers), etc. Terminal device 12 and the organization server 11 may be a single device.

[0047] Figure 4 is a block diagram showing an example configuration of the information processing system 1 according to Embodiment 2. The integrated system 10 comprises the integrated system 10 and a plurality of organizational servers 11, as described in Embodiment 1.

[0048] Furthermore, the organization server 11 includes an output control unit 117 in addition to the functional unit of the organization server 11 in Embodiment 1. Detailed descriptions of each functional unit described in Embodiment 1 are omitted. Here, the analysis unit 116 and the output control unit 117 will be described in detail.

[0049] The analysis unit 116 inputs consumer attribute information into a global model based on acquired parameters and obtains information on consumer consumption trends from the global model. According to the information processing system 1, information on consumer consumption trends that matches the attributes represented by the input attribute information can be obtained.

[0050] The output control unit 117 then outputs the results of the analysis performed by the analysis unit 116. This process of outputting the analysis results includes outputting the results to an external device, such as a terminal device 12. The output format may be audio output, a screen display, or other display output. Furthermore, the output control unit 117 may provide notification based on the analysis results. This notification method may include email or electronic messages.

[0051] One organization can perform analysis based on data held by another organization. For example, it can perform analysis using data that it does not possess. Or, it can perform analysis with higher accuracy than using only the data that one organization possesses. Furthermore, the other company does not need to disclose the data itself to the organization in question, thus maintaining confidentiality. Specifically, in horizontal associative learning, taking customer age groups as an example, data from other customer segments can be used. On the other hand, in vertical associative learning, data on other products for the same customer can be used.

[0052] Furthermore, the analysis unit 116 may analyze, based on purchase history, whether consumers with attributes represented by the input attribute information have purchased the recommended products obtained from the global model. For example, the analysis unit 116 may analyze whether each consumer has purchased the recommended products obtained from the global model. The output control unit 117 may then output the analysis results.

[0053] Next, we will explain the various analyses performed by the analysis unit 116. Specific use cases include supermarkets, apparel stores, publishing houses and bookstores, real estate companies, and banks. Furthermore, we will first explain an example of horizontal associative learning, followed by an example of vertical associative learning.

[0054] <Horizontal Union Learning> First, let's explain horizontal association learning using use cases. We'll use supermarkets, apparel companies, publishing houses, and real estate companies as examples of use cases.

[0055] <Supermarket> The details regarding products and purchase history at supermarkets are as described above.

[0056] For example, let's assume that the main customer base for supermarket A is 50 years old and over, while the main customer base for supermarket B is 20 to 50 years old. For example, supermarket A may want to develop new products for people in their 30s. In the organizational server 11 of supermarket A, the analysis unit 116 inputs attribute information into the global model, for example, where the attribute is age and the attribute value is "30s," and retrieves information on products suitable for people in their 30s from the global model.

[0057] For example, information on products suitable for people in their 30s could be information on products held by supermarket A that are suitable for consumers with the attributes indicated by the attribute information. This allows supermarket A to select products suitable for people in their 30s from its own product lineup based on the information on products for people in their 30s. Alternatively, information on products suitable for people in their 30s could be information on the characteristics of products that are suitable for consumers with the attributes indicated by the attribute information. If the product is food, product characteristics could include flavor, ingredients, and seasonings. For example, a product characteristic suitable for people in their 30s might be a flavor like xxx.

[0058] Furthermore, the output control unit 117 may display the results of the analysis performed by the analysis unit 116 on the display device of the product development terminal device 12. In this way, supermarket A can use the analysis results to develop products for people in their 30s, who are not part of its typical customer base.

[0059] Furthermore, the output control unit 117 may display the results of the analysis by the analysis unit 116 on the display device of the consumer's terminal device 12 corresponding to the attribute indicated by the input attribute information. This allows supermarket A to recommend products suitable for people in their 30s.

[0060] Apparel For example, if the organization is an apparel company, the products are clothing. In such a case, the purchase history would include information about the products purchased by each consumer. The product information could include the product name, type, size, price, and features of each product. For example, features of clothing could include length, color, design, material, and comfort. A feature of clothing might be, for example, "short top."

[0061] Let's assume that apparel company A has previously sold clothing for people in their 40s, but has not sold clothing for people in their 20s. For example, apparel company A has a purchase history for customers in their 40s, but not for customers in their 20s, and therefore faces the challenge of not knowing what products are suitable for customers in their 20s. For example, in apparel company A's organizational server 11, the analysis unit 116 inputs attribute information into the global model, where the attribute is age and the attribute value is "20s," and retrieves information on clothing suitable for people in their 20s from the global model.

[0062] Information on clothing suitable for people in their 20s may be information on clothing items currently owned by apparel company A that are suitable for consumers with the attributes indicated in the attribute information. This allows apparel company A to select products suitable for people in their 20s from its own product lineup based on information on clothing for people in their 20s. Alternatively, information on clothing suitable for people in their 20s may be information on the characteristics of clothing that are suitable for consumers with the attributes indicated in the attribute information.

[0063] The output control unit 117 may then display the results of the analysis by the analysis unit 116 and the attribute information input to the analysis unit 116 on the display device of the product development terminal device 12. In this way, supermarket A can use the analysis results to develop products for people in their 20s, who are not part of its typical customer base.

[0064] Figure 5 is an explanatory diagram showing an example of displaying analysis results on the terminal device 12. For example, the output control unit 117 displays the results of the product development analysis on the display device of the terminal device 12. For example, the terminal device 12 displays the results of the product development analysis according to the control of the output control unit 117.

[0065] In Figure 5, the screen displays recommended features for tops aimed at people in their 20s, based on the results of product development analysis. For example, the screen displays analysis results for each feature, such as color, material, size, and concept.

[0066] Furthermore, the output control unit 117 may display the results of the analysis by the analysis unit 116 on the display device of the consumer's terminal device 12 corresponding to the attribute indicated by the input attribute information. This allows apparel company A to recommend products suitable for consumers in their 20s to consumers in their 20s.

[0067] <Publisher> For example, let's consider an organization with departments such as a newspaper department A, a magazine department B, and a book department C within a publishing company. In such a case, the purchase history would store information about the products purchased by each consumer, including the store where the purchase was made. The store could be a physical store or an e-commerce (EC) website. The product information would include details such as the content of the article or book, the target audience, the price, and the number of pages.

[0068] For example, let's assume that, as a challenge, Book Department C wants to analyze what kind of products should be placed in each bookstore and on e-commerce sites to increase sales. In this case, the model would take the attribute information of the target customers for a new book as input and output information about the stores where customers matching that attribute information would make purchases. Store information could include, for example, the store layout, the arrangement of books in the store, the type of shelves in the store, and the space between products.

[0069] As an example of using the information processing system 1, in the organizational server 11 of the book division C, the analysis unit 116 uses a global model to input attribute information of target consumers for new books and obtains information on stores where consumers matching the attribute information make purchases. The output control unit 117 may then display the attribute information and store information on the display device of the person in charge's terminal device 12. This allows the person in charge of the organization to analyze which bookstores should stock the new books.

[0070] <Real Estate> For example, let's consider the case of an organization such as a real estate company. In the case of real estate, there are rentals and sales, but we will use rentals as an example. Purchase history is, for example, rental history. Rental history includes information such as floor plan, size, rent, and installed equipment for each consumer.

[0071] For example, a real estate company A might want to analyze the floor plans and fixtures of rental properties targeting people in their 20s. A model might then take the attribute information of the target consumer as input and output rental information matching that customer profile. As mentioned above, rental information could include things like floor plans and fixtures.

[0072] As an example of using the information processing system 1, in the organizational server 11 of real estate company A, the analysis unit 116 inputs attribute information, including age and attribute values ​​such as "20s," into the global model and retrieves store information from the global model. The output control unit 117 may then display the attribute information and rental information on the display device of the employee's terminal device 12. This allows the organization's employees to analyze the floor plans and installed equipment of properties rented by consumers in their 20s.

[0073] <Vertical Associative Learning> Next, we will explain vertical associative learning using use cases. We will use supermarkets and apparel stores as examples. While this approach may also be applicable to other use cases such as publishing houses, real estate companies, and financial institutions, we will omit detailed explanations of those.

[0074] <Supermarket> For example, consider a problem where customer X frequently buys fresh produce at supermarket A and prepared foods at supermarket B, but has no purchase history for prepared foods at supermarket A. On the other hand, although customer X buys prepared foods, supermarket A cannot recommend prepared foods to customer X using their purchase history because there is no purchase history for customer X at supermarket A.

[0075] For example, when customer attribute information is input into the model, it outputs information on recommended products suitable for customers whose attributes correspond to those represented by the attribute information. In the organization server 11 of supermarket A, the analysis unit 116 inputs the attribute information of customer X into the global model and obtains information on recommended products for consumers whose attributes correspond to those represented by the input attribute information. Alternatively, for example, the analysis unit 116 may obtain information on recommended products for consumers whose attributes correspond to those represented by the input attribute information from the products sold at supermarket A.

[0076] The output control unit 117 may also cause the customer X's terminal device 12 to display information about recommended products on its display device.

[0077] Figure 6 is an explanatory diagram showing an example of displaying recommended product information on the terminal device 12. In Figure 6, the screen displays "Recommended products for Mr. / Ms. X" from supermarket A. The screen also displays the product name "Prepared food xxxx" and the price "400 yen" as recommended product information. The screen also displays the product name "Prepared food zzzz" and the price "500 yen" as recommended product information.

[0078] Thus, although customer X does not purchase prepared foods at supermarket A, supermarket A can recommend prepared foods to customer X.

[0079] (flowchart) Figure 7 is a flowchart showing an example of the operation of the information processing system 1 according to Embodiment 2. The local model generation unit 111 generates a local model based on the consumer's purchase history and attribute information (step S201). The concealment unit 112 conceals the parameters of the local model using secure computation (step S202). The output unit 113 outputs the concealed parameters to the integrated system 10 (step S203).

[0080] In the integrated system 10, the acquisition unit 101 acquires concealed parameters from each of the multiple organization servers 11 (step S204). The integration unit 102 generates a global model by integrating the concealed parameters of each of the multiple local models (step S205). The provision unit 103 provides the parameters of the global model to the multiple organization servers 11 (step S206).

[0081] In the organizational server 11, the parameter acquisition unit 114 acquires parameters of the global model from the integrated system 10 (step S207). Then, the analysis unit 116 performs analysis using the global model (step S208). The output control unit 117 notifies the terminal device 12 of the analysis results (step S209).

[0082] In Embodiment 2, the consumer trend information is information about recommended products for consumers whose attributes correspond to the attributes represented by the attribute information input into the model. The products are, for example, at least one of the following: food, daily necessities, clothing, books, magazines, and real estate.

[0083] The organization server 11 notifies the terminal device 12 of product recommendations for consumers whose attributes correspond to the attributes represented by the attribute information input into the model. For example, if the notification is sent to the terminal device 12 of an organization's representative, it facilitates the analysis of consumer trends of consumers whose attributes correspond to the attributes represented by the attribute information. Alternatively, it may be used as a product recommendation function, such as notifying consumers' terminal devices 12.

[0084] This concludes the description of each embodiment. The embodiments may be used in combination as appropriate.

[0085] Furthermore, in each embodiment, the integrated system 10 may be configured to include each functional unit and a portion of the information.

[0086] Furthermore, each embodiment is not limited to the examples described above and can be modified in various ways. Also, the configuration of the integrated system 10 in each embodiment is not particularly limited. For example, the integrated system 10 may be implemented by a single device, such as a single server. When each functional part of the integrated system 10 is implemented by a single device, the single device may be called, for example, an integration device, an information processing device, etc., and is not particularly limited. Alternatively, the integrated system in each embodiment may be implemented by different devices for each function or data. For example, each functional part may be composed of multiple servers and implemented as the integrated system 10. For example, the integrated system 10 may be implemented by a database server including each DB (Database) and servers having each functional part. Also, similar to the integrated system 10, the configuration of the organization server 11 in each embodiment is not particularly limited. Furthermore, the organization server 11 may be composed of multiple devices and implemented as an organization system, or it may be composed of a single device.

[0087] Furthermore, in each embodiment, each piece of information and each database may include a portion of the aforementioned information. Also, each piece of information and each database may include information other than the aforementioned information. Each piece of information and each database may be further divided into multiple databases or multiple pieces of information. Thus, the method of implementing each piece of information and each database is not particularly limited. For example, the purchase history may include product identification information that identifies a product, the purchase quantity, and the purchase date, and the product database may store the product identification information that identifies a product and the product information in association. In this way, the purchase history and the product database may be associated by the product identification information that identifies a product.

[0088] Furthermore, each screen is merely an example and is not particularly limited. Buttons, lists, checkboxes, information display fields, input fields, etc., not shown in the illustrations may be added to each screen. Also, the background color of the screen may be changed.

[0089] Furthermore, the process of generating information to be displayed on the terminal device 12 may be performed by the output control unit 117 of the organization server 11. Alternatively, this process may be performed by the terminal device 12.

[0090] (Example of computer hardware configuration) Next, we will describe hardware configuration examples when each of the devices described in each embodiment, such as the integrated system 10, organizational server 11, and terminal device 12, is implemented using a computer. Figure 8 is an explanatory diagram showing an example of computer hardware configuration. Some or all of each device can also be implemented using any combination of computer 80 and program, for example, as shown in Figure 8.

[0091] Computer 80 includes, for example, a processor 801, a ROM (Read Only Memory) 802, a RAM (Random Access Memory) 803, and a storage device 804. Computer 80 also includes a communication interface 805 and an input / output interface 806. Each component is connected, for example, via a bus 807. The number of components is not particularly limited, and each component may be one or more.

[0092] The processor 801 controls the entire computer 80. The processor 801 may include, for example, a CPU (Central Processing Unit), a DSP (Digital Signal Processor), or a GPU (Graphics Processing Unit). The computer 80 has memory units such as ROM 802, RAM 803, and a storage device 804. The storage device 804 may include, for example, semiconductor memory such as flash memory, an HDD (Hard Disk Drive), or an SSD (Solid State Drive). For example, the storage device 804 stores OS (Operating System) programs, application programs, and programs related to each embodiment. Alternatively, ROM 802 stores application programs and programs related to each embodiment. The RAM 803 is used as the work area for the processor 801.

[0093] The processor 801 also loads programs stored in the memory device 804, ROM 802, etc. Then, the processor 801 executes each process coded in the program. The processor 801 may also download various programs via the communication network NT. Furthermore, the processor 801 functions as part or all of the computer 80. The processor 801 may also execute processes or instructions in the illustrated flowchart based on the program.

[0094] The communication interface 805 is connected to a communication network NT, such as a LAN (Local Area Network) or WAN (Wide Area Network), via a wireless or wired communication line. The communication network NT may be composed of multiple communication networks NT. This allows the computer 80 to connect to external devices and external computers 80 via the communication network NT. The communication interface 805 manages the interface between the communication network NT and the internal workings of the computer 80. Furthermore, the communication interface 805 controls the input and output of data from external devices and external computers 80.

[0095] Furthermore, the input / output interface 806 is connected to at least one of the input device, output device, and input / output device. The connection method may be wireless or wired. Examples of input devices include keyboards, mice, and microphones. Examples of output devices include display devices, lighting devices, and audio output devices. Examples of input / output devices include touch panel displays. The input device, output device, and input / output device may be built into the computer 80 or may be external.

[0096] The hardware configuration of computer 80 is an example. Computer 80 may have some of the components shown in Figure 8. Computer 80 may have components other than those shown in Figure 8. For example, computer 80 may have a drive device. The processor 801 may read programs and data stored on a recording medium attached to the drive device into RAM 803. Examples of non-temporary tangible recording media include optical discs, flexible discs, magneto-optical discs, and USB (Universal Serial Bus) memory. Also, as mentioned above, computer 80 may have input devices such as a keyboard and a mouse. Computer 80 may have output devices such as a display. Furthermore, computer 80 may have input devices, output devices, and input / output devices, respectively.

[0097] Furthermore, the computer 80 may have various sensors (not shown). The type of sensor is not particularly limited. Also, the computer 80 may be equipped with an imaging device capable of capturing images or videos.

[0098] This concludes the description of the hardware configuration of each device. Furthermore, there are various variations in how each device can be implemented. For example, each device may be implemented by any combination of different computers and programs for each component. Alternatively, the multiple components of each device may be implemented by any combination of a single computer and program.

[0099] Furthermore, some or all of the components of each device may be implemented by application-specific circuits. Alternatively, some or all of the components of each device may be implemented by general-purpose circuits, including processors such as FPGAs (Field Programmable Gate Arrays). Furthermore, some or all of the components of each device may be implemented by a combination of application-specific circuits and general-purpose circuits. These circuits may also be a single integrated circuit, or they may be divided into multiple integrated circuits. These multiple integrated circuits may be connected via a bus or the like.

[0100] Furthermore, if some or all of the components of each device are implemented by multiple computers or circuits, these computers or circuits may be centrally located or distributed.

[0101] The integration method described in each embodiment is implemented by the integration system 10. Alternatively, for example, the integration method can be implemented by a computer such as a server or terminal device executing a pre-prepared program. Furthermore, the analysis method described in each embodiment is implemented by the organization server. Alternatively, for example, the analysis method can be implemented by a computer such as a server or terminal device executing a pre-prepared program.

[0102] The programs described in each embodiment are recorded on a computer-readable recording medium such as an HDD, SSD, flexible disk, optical disk, magneto-optical disk, or USB memory. The programs are then executed by being read from the recording medium by a computer. The programs may also be distributed via a communication network NT.

[0103] Each component of the integrated system 10 in each embodiment described above may be implemented using dedicated hardware, such as a computer. Alternatively, each component may be implemented using software. Alternatively, each component may be implemented using a combination of hardware and software. Similarly, each component of the organizational server may be implemented in the same way.

[0104] The present disclosure has been described above with reference to the embodiments described herein, but the present disclosure is not limited to the embodiments described above. The structure and details of each present disclosure may include embodiments that apply various modifications that can be grasped by those skilled in the art within the scope of the present disclosure. The present disclosure may include embodiments that combine or substitute the matters described herein as appropriate. For example, matters described using a particular embodiment may be applied to other embodiments to the extent that they do not cause a contradiction. For example, although multiple operations are described sequentially in the form of a flowchart, the order in which they are described does not limit the order in which the multiple operations are performed. Therefore, when implementing each embodiment, the order of the multiple operations may be changed to the extent that it does not impair the content.

[0105] Some or all of the above embodiments may also be described as follows. However, some or all of the above embodiments are not limited to the following.

[0106] (Note 1) For each of the multiple organizational servers, an acquisition means for acquiring anonymized parameters of a local model that has been learned based on the consumer purchase history and attribute information of the consumer held by each server, and which, when attribute information is input, can output information regarding the consumption trends of the consumer corresponding to the attribute represented by the input attribute information. An integration means for generating a global model by integrating confidential parameters for each of the aforementioned multiple organizational servers, A means for providing the integrated parameters of the generated global model to the plurality of organizational servers, An integrated system equipped with [the following features]. (Note 2) The integration means selects some confidential parameters from the confidential parameters for each of the plurality of organization servers based on the characteristics of each of the plurality of organization servers, and integrates the selected confidential parameters. The providing means provides the integrated parameters to one of the multiple organizational servers that is the source of the confidential parameters. The integrated system described in Appendix 1. (Note 3) The aforementioned consumer trend information is information on recommended products for the consumer that correspond to the attribute represented by the input attribute information. The integrated system described in Appendix 1 or 2. (Note 4) The aforementioned goods are at least one of the following: food, daily necessities, clothing, books, magazines, or real estate. The integrated system described in Appendix 3. (Note 5) A model generation means that learns a local model that, based on a consumer's purchase history and attribute information, can output information about the consumer's consumption trends corresponding to the attribute represented by the input attribute information when attribute information is input. The parameters of the local model are provided as an anonymization means for concealing them using secure computation, An output means for outputting the confidential parameters of the local model to the integrated system, A parameter acquisition means for acquiring confidential parameters of a global model integrated in the integrated system from the integrated system, An analysis means that inputs attribute information into the global model based on the acquired parameters, and obtains information from the global model regarding the consumption trends of consumers corresponding to the attributes represented by the input attribute information, Organizational servers equipped with these features. (Note 6) An output control means that notifies the acquired information regarding consumption trends, The organizational server described in Appendix 5, which includes the following features. (Note 7) For each of the multiple organizational servers, a local model is trained based on the consumer purchase history and attribute information of the consumer, and when attribute information is input, the local model's confidential parameters are obtained, which can output information about the consumer's consumption trends corresponding to the attribute represented by the input attribute information. The confidential parameters for each of the aforementioned multiple organizational servers are integrated, To provide integrated parameters to the aforementioned multiple organizational servers, A global model is generated by integrating the confidential parameters for each of the aforementioned multiple organizational servers. The integrated parameters of the generated global model are provided to the aforementioned multiple organizational servers. Integration method. (Note 8) Based on a consumer's purchase history and attribute information, when attribute information is input, a local model is trained that can output information about the consumer's consumption trends corresponding to the attribute represented by the input attribute information. The parameters of the aforementioned local model are made secret using secure computation. The confidential parameters of the local model are output to the integrated system. From the aforementioned integrated system, obtain the parameters of the global model integrated in the aforementioned integrated system. The global model, based on the acquired parameters, is populated with attribute information, and information regarding consumer spending trends corresponding to the attributes represented by the input attribute information is obtained from the global model. Analysis method. (Note 9) On the computer, For each of the multiple organizational servers, a local model is trained based on the consumer purchase history and attribute information of the consumer, and when attribute information is input, the local model's confidential parameters are obtained, which can output information about the consumer's consumption trends corresponding to the attribute represented by the input attribute information. A global model is generated by integrating the confidential parameters for each of the aforementioned multiple organizational servers. The integrated parameters of the generated global model are provided to the aforementioned multiple organizational servers. A non-temporary recording medium readable by the computer, which records a program that executes a process. (Note 10) On the computer, Based on a consumer's purchase history and attribute information, when attribute information is input, a local model is trained that can output information about the consumer's consumption trends corresponding to the attribute represented by the input attribute information. The parameters of the aforementioned local model are made secret using secure computation. The confidential parameters of the local model are output to the integrated system. From the aforementioned integrated system, obtain the parameters of the global model integrated in the aforementioned integrated system. The global model, based on the acquired parameters, is populated with attribute information, and information regarding consumer spending trends corresponding to the attributes represented by the input attribute information is obtained from the global model. A non-temporary recording medium readable by the computer, which records a program that executes a process. (Note 11) On the computer, For each of the multiple organizational servers, a local model is trained based on the consumer purchase history and attribute information of the consumer, and when attribute information is input, the local model's confidential parameters are obtained, which can output information about the consumer's consumption trends corresponding to the attribute represented by the input attribute information. A global model is generated by integrating the confidential parameters for each of the aforementioned multiple organizational servers. The integrated parameters of the generated global model are provided to the aforementioned multiple organizational servers. A program that executes a process. (Note 12) On the computer, Based on a consumer's purchase history and attribute information, when attribute information is input, a local model is trained that can output information about the consumer's consumption trends corresponding to the attribute represented by the input attribute information. The parameters of the aforementioned local model are made secret using secure computation. The confidential parameters of the local model are output to the integrated system. From the aforementioned integrated system, obtain the parameters of the global model integrated in the aforementioned integrated system. The global model, based on the acquired parameters, is populated with attribute information, and information regarding consumer spending trends corresponding to the attributes represented by the input attribute information is obtained from the global model. A program that executes a process. [Explanation of symbols]

[0107] 1. Information Processing System 10. Integrated System 11,11a,11b Organization Server 12 Terminal devices 80 Computers 101 Acquisition Department 102 Integration Department 103 Provision Department 111 Local Model Generation Unit 112 Confidentiality Section 113 Output section 114 Parameter acquisition unit 115 Model Memory Unit 116 Analysis Department 117 Output Control Unit 801 Processor 802 ROM 803 RAM 804 Storage device 805 Communication Interface 806 Input / Output Interface 807 Bus NT communication network

Claims

1. For each of the multiple organizational servers, an acquisition means for acquiring anonymized parameters of a local model that has been learned based on the consumer purchase history and attribute information of the consumer held by each server, and which, when attribute information is input, can output information regarding the consumption trends of the consumer corresponding to the attribute represented by the input attribute information. An integration means for generating a global model by integrating confidential parameters for each of the aforementioned multiple organizational servers, A means for providing the integrated parameters of the generated global model to the plurality of organizational servers, Equipped with, Each of the aforementioned multiple organizational servers learns the local model based on the consumer's purchase history and attribute information it possesses, and conceals the parameters of the local model using secure computation. The acquisition means receives anonymized parameters from each of the plurality of organizational servers. Integrated system.

2. The integration means selects some confidential parameters from the confidential parameters for each of the plurality of organization servers based on the characteristics of each of the plurality of organization servers, and integrates the selected confidential parameters. The providing means provides the integrated parameters to one of the multiple organizational servers that is the source of the confidential parameters. The integrated system according to claim 1.

3. The aforementioned consumer trend information is information on recommended products for the consumer that correspond to the attribute represented by the input attribute information. The integrated system according to claim 1 or 2.

4. The aforementioned goods are at least one of the following: food, daily necessities, clothing, books, magazines, or real estate. The integrated system according to claim 3.

5. A model generation means that learns a local model that, based on a consumer's purchase history and attribute information, can output information about the consumer's consumption trends corresponding to the attribute represented by the input attribute information when attribute information is input. The parameters of the local model are provided as an anonymization means for concealing them using secure computation, An output means for outputting the confidential parameters of the local model to the integrated system, A parameter acquisition means for acquiring parameters of a global model integrated in the integrated system from the integrated system, An analysis means that inputs attribute information into the global model based on the acquired parameters, and obtains information from the global model regarding the consumption trends of consumers corresponding to the attributes represented by the input attribute information, Equipped with, The aforementioned integration system generates the global model by integrating the anonymized parameters output from each of several organizational servers, including its own organizational server. The parameter acquisition means receives the parameters of the integrated global model from the integrated system. Organizational server.

6. An output control means that notifies the acquired information regarding consumption trends, The organizational server according to claim 5, comprising:

7. The computer of the integrated system, For each of the multiple organizational servers, a local model is trained based on the consumer purchase history and attribute information of the consumer, and when attribute information is input, the local model's confidential parameters are obtained, which can output information about the consumer's consumption trends corresponding to the attribute represented by the input attribute information. A global model is generated by integrating the confidential parameters for each of the aforementioned multiple organizational servers. The integrated parameters of the generated global model are provided to the aforementioned multiple organizational servers. Execute the process, Each of the aforementioned multiple organizational servers learns the local model based on the consumer's purchase history and attribute information it possesses, and conceals the parameters of the local model using secure computation. In the process of obtaining the confidential parameters of the local model, the confidential parameters are received from each of the multiple organization servers. Integration method.

8. The computer of the organization server, Based on a consumer's purchase history and attribute information, when attribute information is input, a local model is trained that can output information about the consumer's consumption trends corresponding to the attribute represented by the input attribute information. The parameters of the aforementioned local model are made secret using secure computation. The confidential parameters of the local model are output to the integrated system. From the aforementioned integrated system, obtain the parameters of the global model integrated in the aforementioned integrated system. The global model, based on the acquired parameters, is populated with attribute information, and information regarding consumer spending trends corresponding to the attributes represented by the input attribute information is obtained from the global model. Execute the process, The aforementioned integration system generates the global model by integrating the anonymized parameters output from each of several organizational servers, including its own organizational server. In the process of obtaining the parameters of the global model, the integrated parameters of the global model are received from the integrated system. Analysis method.

9. The computer of the integrated system, For each of the multiple organizational servers, a local model is trained based on the consumer purchase history and attribute information of the consumer, and when attribute information is input, the local model's confidential parameters are obtained, which can output information about the consumer's consumption trends corresponding to the attribute represented by the input attribute information. A global model is generated by integrating the confidential parameters for each of the aforementioned multiple organizational servers. The integrated parameters of the generated global model are provided to the aforementioned multiple organizational servers. Execute the process, Each of the aforementioned multiple organizational servers learns the local model based on the consumer's purchase history and attribute information it possesses, and conceals the parameters of the local model using secure computation. In the process of obtaining the confidential parameters of the local model, the confidential parameters are received from each of the multiple organization servers. program.

10. The computer of the organizational server, Based on a consumer's purchase history and attribute information, when attribute information is input, a local model is trained that can output information about the consumer's consumption trends corresponding to the attribute represented by the input attribute information. The parameters of the aforementioned local model are made secret using secure computation. The confidential parameters of the local model are output to the integrated system. From the aforementioned integrated system, obtain the parameters of the global model integrated in the aforementioned integrated system. The global model, based on the acquired parameters, is populated with attribute information, and information regarding consumer spending trends corresponding to the attributes represented by the input attribute information is obtained from the global model. Execute the process, The aforementioned integration system generates the global model by integrating the anonymized parameters output from each of several organizational servers, including its own organizational server. In the process of obtaining the parameters of the global model, the integrated parameters of the global model are received from the integrated system. program.

Images