Information processing device, information processing method, and recording medium

Abstract

In this information processing device, a training means trains a learning model using first graph data indicating attributes of conventional products, second graph data indicating attributes of customers who visit a prescribed store, third graph data indicating the purchase status of the conventional products, and relationship data, which is data indicating known relationships between nodes linked in the first graph data, the second graph data, and the third graph data. A purchase prediction means uses the trained learning model to predict the purchase of a new product for each customer who visits the prescribed store. A demand prediction means uses the result of the purchase prediction to predict the demand for the new product at the prescribed store. A display data output means outputs data relating to the result of the demand prediction as display data.

Description

Information processing device, information processing method, and recording medium

[0001] The present disclosure relates to a technique that can be used to forecast demand for a product.

[0002] Techniques related to product demand forecasting have been proposed.

[0003] Specifically, for example, Patent Document 1 discloses a technology that generates integrated behavioral history information for each person by integrating cyber behavioral history information related to past behavior on the Internet and real behavioral history information related to past behavior in a physical store, and analyzes and processes the integrated behavioral history information to generate purchase prediction information related to products that target customers are predicted to purchase.

[0004] International Publication No. WO2020 / 008938

[0005] However, the technology disclosed in Patent Document 1 has the problem that it is difficult to predict the purchase of newly released products, for example, because there is no purchase history of the products in physical stores.

[0006] Therefore, the technology disclosed in Patent Document 1 has a problem corresponding to the above-mentioned problem that it is not possible to obtain information useful for predicting demand for newly released products.

[0007] One object of the present disclosure is to provide an information processing device capable of obtaining information useful for predicting demand for newly released products.

[0008] In one aspect of the present disclosure, an information processing device includes: a learning means for learning a learning model using first graph data indicating the attributes of conventional products, second graph data indicating the attributes of customers visiting a specified store, third graph data indicating the purchasing status of the conventional products, and relationship data which is data indicating known relationships between nodes linked in the first graph data, the second graph data, and the third graph data; a purchase prediction means for predicting purchases of new products for each customer visiting the specified store using the learned learning model; a demand prediction means for predicting demand for the new products at the specified store using the prediction results of the purchase prediction; and a display data output means for outputting data related to the prediction results of the demand prediction as display data.

[0009] In another aspect of the present disclosure, an information processing method is an information processing method executed by a computer, which trains a learning model using first graph data indicating attributes of conventional products, second graph data indicating attributes of customers visiting a specified store, third graph data indicating the purchase status of the conventional products, and relationship data that is data indicating known relationships between nodes linked in the first graph data, the second graph data, and the third graph data, and uses the trained learning model to make a purchase prediction of a new product for each customer visiting the specified store, uses the prediction results of the purchase prediction to make a demand prediction for the new product at the specified store, and outputs data related to the prediction results of the demand forecast as display data.

[0010] In yet another aspect of the present disclosure, a recording medium records a program that causes a computer to execute the following process: learning a learning model using first graph data indicating the attributes of conventional products, second graph data indicating the attributes of customers visiting a specified store, third graph data indicating the purchasing status of the conventional products, and relationship data that is data indicating known relationships between nodes linked in the first graph data, the second graph data, and the third graph data; using the learned learning model, making a purchase prediction of a new product for each customer visiting the specified store; using the prediction results of the purchase prediction, making a demand prediction for the new product at the specified store; and outputting data related to the prediction results of the demand forecast as display data.

[0011] According to the present disclosure, it is possible to obtain information useful for predicting demand for newly released products.

[0012] 1 is a diagram showing a schematic configuration of an information processing system including a server device according to the present disclosure. FIG. 1 is a block diagram showing an example of a hardware configuration of a server device according to the present disclosure. FIG. 2 is a block diagram showing an example of a functional configuration of a server device according to the present disclosure. FIG. 3 is a diagram showing an example of conventional product data used in processing of a server device according to the present disclosure. FIG. 4 is a diagram showing an example of customer data used in processing of a server device according to the present disclosure. FIG. 5 is a diagram showing an example of purchase data used in processing of a server device according to the present disclosure. FIG. 6 is a diagram showing an example of new product data used in processing of a server device according to the present disclosure. FIG. 7 is a diagram showing an example of example data acquired by processing of a server device according to the present disclosure. FIG. 8 is a diagram showing an example of an operation screen displayed in response to processing of a server device according to the present disclosure. FIG. 9 is a diagram showing another example of an operation screen displayed in response to processing of a server device according to the present disclosure. FIG. 10 is a diagram showing another example of an operation screen displayed in response to processing of a server device according to the present disclosure. FIG. 11 is a flowchart showing an example of processing performed in a server device according to the present disclosure. FIG. 12 is a block diagram showing an example of a functional configuration of an information processing device according to the present disclosure. FIG. 13 is a flowchart showing an example of processing performed in an information processing device according to the present disclosure.

[0013] Hereinafter, preferred embodiments of the present disclosure will be described with reference to the drawings.

[0014] First Embodiment [System Configuration] Fig. 1 is a diagram showing a schematic configuration of an information processing system including a server device according to the present disclosure. As shown in Fig. 1, the information processing system 1 includes a server device 100 and a terminal device 200.

[0015] The server device 100 is configured to be able to communicate with the terminal device 200. The server device 100 also performs processing related to, for example, demand forecasting for newly released products (hereinafter also referred to as new products). In response to an instruction signal input from the terminal device 200, the server device 100 outputs data related to the results of the demand forecasting for the new products to the terminal device 200 as display data.

[0016] The terminal device 200 has a function of communicating with the server device 100, a function of inputting information to be transmitted to the server device 100, and a function of displaying information received from the server device 100. The terminal device 200 also has a function of displaying information in response to a user operation. Specifically, the terminal device 200 may be configured by a device such as a personal computer, a smartphone, or a tablet computer, for example.

[0017] [Hardware Configuration] Fig. 2 is a block diagram showing an example of the hardware configuration of a server device according to the present disclosure. As shown in Fig. 2, the server device 100 includes an interface (IF) 111, a processor 112, a memory 113, a recording medium 114, and a database (DB) 115.

[0018] The IF 111 inputs and outputs data to and from external devices. For example, information transmitted from the terminal device 200 is input to the server device 100 via the IF 111.

[0019] The processor 112 is a computer such as a CPU (Central Processing Unit), and executes a prepared program to control the entire server device 100. Specifically, the processor 112 performs analysis using, for example, a learning model described below.

[0020] The memory 113 is configured by a ROM (Read Only Memory), a RAM (Random Access Memory), etc. The memory 113 is also used as a working memory while the processor 112 is executing various processes.

[0021] The recording medium 114 is a non-volatile, non-transitory recording medium such as a disk-shaped recording medium or a semiconductor memory, and is configured to be detachable from the server device 100. The recording medium 114 records various programs to be executed by the processor 112. When the server device 100 executes various processes, the programs recorded on the recording medium 114 are loaded into the memory 113 and executed by the processor 112.

[0022] The DB 115 stores, for example, information input via the IF 111 and processing results obtained by processing by the processor 112 .

[0023] [Functional Configuration] Fig. 3 is a block diagram showing an example of the functional configuration of a server device according to the present disclosure. As shown in Fig. 3, the server device 100 includes a graph data storage unit 11, a learning processing unit 12, a processing result storage unit 13, a purchase prediction processing unit 15, and a display processing unit 16.

[0024] The graph data storage unit 11 stores conventional product data EGD, customer data CGD, and purchase data PGD as a plurality of graph data.

[0025] The conventional product data EGD is created as graph data showing attributes of products (hereinafter also referred to as conventional products) that were released before the new product, for example, as shown in Figure 4. Furthermore, the conventional product data EGD is created as data in which information related to the connection source node is represented in a "head" column, information related to the connection destination node is represented in a "tail" column, and information related to the connection relationship from the "head" to the "tail" column is represented. Figure 4 is a diagram showing an example of conventional product data used in processing by a server device according to the present disclosure.

[0026] The "head" column of the conventional product data EGD contains a character string, such as "SKU_001," that allows individual conventional products to be identified. The "tail" column of the conventional product data EGD contains a character string that indicates the attribute of the conventional product, such as "white." The "relation" column of the conventional product data EGD contains a character string that indicates the category of the attribute of the conventional product, such as "representative color." Therefore, according to the conventional product data EGD of FIG. 4, it is possible to identify that the representative color of the conventional product corresponding to "SKU_001," for example, is white.

[0027] The customer data CGD is created as graph data showing the attributes of customers visiting a specific store PS, as shown in FIG. 5 . Customers visiting a specific store PS may include, for example, individuals registered as customers of the specific store PS and individuals with a history of purchasing products at the specific store PS. The customer data CGD can be created using registration information previously registered by the customer. It is desirable that the registration information includes, for example, information that identifies one of multiple stores as the specific store PS. The customer data CGD is created as data in which information related to the connection source node is represented in the “head” column, information related to the connection destination node is represented in the “tail” column, and information related to the connection relationship from the “head” to the “tail” column is represented in the “relation” column. FIG. 5 is a diagram illustrating an example of customer data used in processing by a server device according to the present disclosure.

[0028] The "head" column of the customer data CGD contains a character string, such as "customer001," that allows individual customers to be identified. The "tail" column of the customer data CGD contains a character string that indicates the attributes of the customer, such as "Male." The "relation" column of the customer data CGD contains a character string that indicates the category of the attributes possessed by the customer, such as "gender." Therefore, according to the customer data CGD of FIG. 5, it is possible to identify, for example, that the gender of the customer corresponding to "customer001" is male.

[0029] The purchase data PGD is created as graph data showing the purchase status of conventional products, for example, as shown in Fig. 6. The purchase data PGD is created as data in which information related to the connection source node is represented in a "head" column, information related to the connection destination node is represented in a "tail" column, and information related to the connection relationship from the "head" to the "tail" column is represented in a "relation" column. Fig. 6 is a diagram showing an example of purchase data used in processing by a server device according to the present disclosure.

[0030] The "head" column of the purchase data PGD contains a character string, such as "customer001," that identifies the customer who purchased the conventional product. The "tail" column of the purchase data PGD contains a character string, such as "SKU_001," that identifies the conventional product purchased by the customer. The "relation" column of the purchase data PGD contains a character string, such as "purchase," that indicates that the conventional product has been purchased. Therefore, according to the purchase data PGD of FIG. 6, it is possible to identify that the customer corresponding to "customer001" purchased the conventional product corresponding to "SKU_001."

[0031] For ease of explanation, hereinafter, the conventional product data EGD will be referred to as graph data EGD, the customer data CGD will be referred to as graph data EGD, and the purchase data PGD will be referred to as graph data PGD.

[0032] The learning processing unit 12 functions as a learning means. The learning processing unit 12 acquires, as relationship data KD, data indicating known relationships between linked nodes in the three graph data EGD, CGD, and PGD read from the graph data storage unit 11. The relationship data KD includes data indicating relationships between multiple nodes as universal rules by replacing each of the multiple nodes having the same relationship with a variable node. The learning processing unit 12 also uses the three graph data EGD, CGD, and PGD and the relationship data KD to train the learning model GMD so as to derive unknown relationships between unlinked nodes in the three graph data. Specifically, the learning processing unit 12 trains the learning model GMD by inputting, for example, feature values ​​corresponding to each node included in the three graph data EGD, CGD, and PGD and feature values ​​corresponding to each relationship included in the relationship data KD into the learning model GMD constructed based on "KBLRN." In addition, the learning processing unit 12 stores the relationship data KD and the learned learning model GMD in the processing result storage unit 13.

[0033] The aforementioned "KBLRN" is disclosed, for example, in "KBLRN: End-to-End Learning of Knowledge-Based Representations with Latent, Relational, and Numerical Features" by Alberto Garcia-Duran et al. Furthermore, the learning model GMD may be constructed based on a model other than "KBLRN" as long as it has a configuration capable of performing link prediction for graph data. Furthermore, when the learning processing unit 12 learns the learning model GMD using the three graph data EGD, CGD, and PGD and the relationship data KD, it is desirable to perform zero-shot learning, such as that disclosed in Japanese Patent Application Laid-Open No. 2019-125364. Furthermore, according to this embodiment, for example, each time at least one of the three graph data EGD, CGD, and PGD is updated, the learning model GMD may be re-learned using the updated graph data.

[0034] The processing result storage unit 13 stores the relationship data KD and the learned learning model GMD as data obtained by the processing of the learning processing unit 12.

[0035] The purchase prediction processing unit 15 functions as a purchase prediction unit. Furthermore, the purchase prediction processing unit 15 uses the trained learning model GMD read from the processing result storage unit 13 to perform processing related to purchase prediction of a new product corresponding to the new product data NGD input from the terminal device 200. Specifically, the purchase prediction processing unit 15 calculates, for each customer visiting a specific store PS, a score SC corresponding to the purchase probability of the new product corresponding to the new product data NGD, and acquires the calculated score SC for each customer as a prediction result PRC of the purchase prediction of the new product. That is, the purchase prediction processing unit 15 can use the trained learning model GMD to predict the purchase of a new product for each customer visiting a specific store PS. Furthermore, the purchase prediction processing unit 15 can acquire, as the prediction result PRC of the purchase prediction, a score SC corresponding to the purchase probability of the new product for each customer visiting a specific store PS. Furthermore, the purchase prediction processing unit 15 outputs the new product data NGD and the prediction result PRC for each customer related to the new product data NGD to the display processing unit 16.

[0036] The new product data NGD is created as graph data showing attributes of new products, for example, as shown in Fig. 7. The new product data NGD is created as data in which information related to the connection source node is represented in a "head" column, information related to the connection destination node is represented in a "tail" column, and information related to the connection relationship from the "head" to the "tail" column is represented in a "relation" column. Fig. 7 is a diagram showing an example of new product data used in processing by the server device according to the present disclosure.

[0037] The "head" column of the new product data NGD contains a character string, such as "NKU_003," that allows each new product to be identified. The "tail" column of the new product data NGD contains a character string that indicates the attribute of the new product, such as "white." The "relation" column of the new product data NGD contains a character string that indicates the category of the attribute of the new product, such as "representative color." Therefore, according to the new product data NGD in FIG. 7 , it is possible to identify that the representative color of the new product corresponding to "NKU_003" is white. It is desirable that the new product data NGD contain the same type of information as the information contained in the conventional product data EGD.

[0038] The display processing unit 16 acquires data related to the prediction results of the demand forecast for new products in a predetermined store PS, using the three graph data EGD, CGD, and PGD read from the graph data storage unit 11, the relationship data KD read from the processing result storage unit 13, the new product data NGD and prediction results PRC obtained from the purchase prediction processing unit 15, and the store data SD input from the terminal device 200. Furthermore, in response to an instruction signal SS input from the terminal device 200, the display processing unit 16 outputs data necessary for generating an operation screen including a GUI (Graphical User Interface) related to the demand forecast for new products to the terminal device 200 as display data HD.

[0039] The store data SD may include at least one piece of data indicating a situation specific to a predetermined store PS. Specifically, the store data SD may include, as data indicating a situation specific to a predetermined store PS, at least one piece of data such as an upper limit on the number of products that can be purchased with a predetermined attribute, a budget that can be allocated to new products, and a purchase quantity adjustment value that is applied to products with a predetermined attribute.

[0040] The display processing unit 16 includes a rule extraction unit 16A, a demand forecast processing unit 16B, an example data acquisition unit 16C, and a display control unit 16D.

[0041] The rule extraction unit 16A extracts a rule RL that is the basis for the prediction result PRC for each customer from among the rules included in the relationship data KD read from the processing result storage unit 13.

[0042] The demand forecast processing unit 16B functions as a demand forecasting means. The demand forecast processing unit 16B performs processing related to the demand forecast for new products at a specific store PS using the forecast results PRC, store data SD, and new product data NGD. The demand forecast processing unit 16B also includes a recommended purchase quantity acquisition unit 16P.

[0043] Here, a specific example of the processing performed by the demand forecast processing unit 16B will be described.

[0044] The demand forecasting processing unit 16B uses the store data SD and the new product data NGD to set extraction criteria EC for extracting customers likely to purchase new products from the prediction result PRC. Specifically, for example, if a new product corresponding to the new product data NGD has a predetermined attribute and an upper limit UA for the purchase quantity of a product having the predetermined attribute is included in the store data SD, the demand forecasting processing unit 16B can set the number of customers corresponding to the upper limit UA as the extraction criteria EC. Furthermore, for example, if the attributes of the new product data NGD include the sales price of the new product and the store data SD includes a budget that can be allocated to the new product, the demand forecasting processing unit 16B can calculate the upper limit UA for the purchase quantity by dividing the budget by the sales price, and set the number of customers corresponding to the upper limit UA as the extraction criteria EC.

[0045] The demand forecasting processing unit 16B extracts customers having a score SC equal to or greater than a predetermined score based on the prediction result PRC for each customer. The demand forecasting processing unit 16B also extracts customers corresponding to the extraction criteria EC from each customer having a score SC equal to or greater than the predetermined score. Specifically, if an upper limit UA is set as the extraction criteria EC, the demand forecasting processing unit 16B extracts customers having a score SC equal to or greater than the predetermined score, in descending order of their scores SC, with the number of customers equal to or less than the upper limit UA. Hereinafter, for ease of explanation, a group including at least one customer extracted by the demand forecasting processing unit 16B from the prediction result PRC based on the score SC and the extraction criteria EC will be referred to as a customer group CKG. That is, the customer group CKG includes at least one customer predicted to purchase a new product. The demand forecasting processing unit 16B can also extract, from each customer having a score equal to or greater than a predetermined score in the prediction result PRC for each customer, customers corresponding to the extraction criteria EC set based on the circumstances specific to a specific store PS and the attributes of a new product as customers belonging to the customer group CKG.

[0046] The recommended purchase quantity acquisition unit 16P acquires the number of customers belonging to the customer group CKG as the recommended purchase quantity SPN for the new product. For example, if a new product corresponding to the new product data NGD has a predetermined attribute and the store data SD includes a purchase quantity correction value applied to products with the predetermined attribute, the recommended purchase quantity acquisition unit 16P can acquire the value obtained by multiplying the number of customers belonging to the customer group CKG by the correction value as the recommended purchase quantity SPN for the new product. Specifically, for example, if the correction value is "2.0," the recommended purchase quantity acquisition unit 16P can acquire a value that is twice the number of customers belonging to the customer group CKG as the recommended purchase quantity SPN.

[0047] According to the above-described process, the demand forecast processing unit 16B can obtain the customer group CKG and the recommended purchase quantity SPN as the forecast result PRD of the demand forecast for a new product in a predetermined store PS.

[0048] The example data acquisition unit 16C extracts a rule RLG corresponding to a customer belonging to the customer group CKG of the prediction result PRD from each rule RL extracted as the basis of the prediction result PRC. The example data acquisition unit 16C also uses the three graph data EGD, CGD, and PGD read from the graph data storage unit 11 to generate graph data showing past purchase cases in which the rule RLG holds, and acquires the graph data as example data JGD. The example data acquisition unit 16C also acquires example data JGD for each customer belonging to the customer group CKG of the prediction result PRD. Specifically, the example data acquisition unit 16C can acquire example data JGDA, such as that shown in FIG. 8, as example data for a customer corresponding to "customer001" belonging to the customer group CKG. The example data JGDA is acquired as graph data showing the basis for the prediction result that a customer corresponding to "customer001" belonging to customer group CKG will purchase a new product corresponding to "NKU_003." The example data JGDA also shows, as the basis for the prediction result, that the conventional product corresponding to "SKU_001" purchased by "customer001" and the new product corresponding to "NKU_003" predicted to be purchased by "customer001," belong to the same brand and have the same representative color. Figure 8 is a diagram showing an example of example data acquired by processing by a server device according to the present disclosure.

[0049] The display control unit 16D functions as a display data output unit. The display control unit 16D also references the new product data NGD to acquire new product list data NLD, which includes character strings for individually identifying the new products included in the new product data NGD. The display control unit 16D also outputs data corresponding to an instruction signal SS input from the terminal device 200 to the terminal device 200 as display data HD.

[0050] Here, a description will be given of a process performed by the display control unit 16D and a specific example of an operation screen displayed on the terminal device 200 in response to the process.

[0051] For example, when an instruction signal SS for starting a demand forecast for new products at a predetermined store PS is input from the terminal device 200, the display control unit 16D outputs the new product list data NLD as display data HD to the terminal device 200. According to this processing, for example, the operation screen SGA of FIG. 9 and the operation screen SGB of FIG. 10 can be displayed on the terminal device 200. FIG. 9 is a diagram showing an example of an operation screen displayed in response to processing by the server device according to the present disclosure. FIG. 10 is a diagram showing another example of an operation screen displayed in response to processing by the server device according to the present disclosure.

[0052] The operation screen SGA includes a selection box NSB, which corresponds to a GUI for selecting a new product. The operation screen SGB includes the selection box NSB and a pull-down list NPD. The pull-down list NPD displays items corresponding to each new product included in the new product list data NLD, such as "NKU_003" and "NKU_004."

[0053] For example, the user can press the selection box NSB on the operation screen SGA to display the pull-down list NPD on the operation screen SGB, and can select one item from the items displayed in the pull-down list NPD.

[0054] The terminal device 200 outputs an instruction signal SS indicating that one item in the pull-down list NPD has been selected to the server device 100. Specifically, the terminal device 200 outputs an instruction signal SS indicating that, for example, "NKU_003" in the pull-down list NPD has been selected to the server device 100.

[0055] When an instruction signal SS indicating that "NKU_003" has been selected is input from the terminal device 200, the display control unit 16D extracts data indicating the attributes of "NKU_003" from the new product data NGD and acquires the extracted data as attribute data NZD. Furthermore, when an instruction signal SS indicating that "NKU_003" has been selected is input from the terminal device 200, the display control unit 16D acquires the number of customers belonging to the customer group CKG corresponding to "NKU_003" as predicted purchase customer number data PCD. Furthermore, when an instruction signal SS indicating that "NKU_003" has been selected is input from the terminal device 200, the display control unit 16D refers to the graph data CGD to acquire customer list data CLD including character strings for individually identifying customers belonging to the customer group CKG corresponding to "NKU_003." That is, the customer list data CLD indicates customers belonging to the customer group CKG who are predicted to purchase the new product corresponding to "NKU_003." Furthermore, when an instruction signal SS indicating that "NKU_003" has been selected is input from the terminal device 200, the display control unit 16D acquires the data on the recommended purchase quantity SPN corresponding to "NKU_003" as the recommended purchase quantity data SPND. The display control unit 16D then outputs the attribute data NZD, the predicted purchase customer number data PCD, the customer list data CLD, and the recommended purchase quantity data SPND to the terminal device 200 as display data HD. According to this processing, for example, the operation screen SGC of FIG. 11 and the operation screen SGD of FIG. 12 can be displayed on the terminal device 200. FIGS. 11 and 12 are diagrams showing other examples of operation screens displayed in response to processing by the server device according to the present disclosure.

[0056] The operation screen SGC includes a selection box NSB with "NKU_003" selected. The operation screen SGC also includes attribute information NZJ indicating the attributes of "NKU_003" included in the attribute data NZD, predicted customer number of purchase information PCJ corresponding to the predicted customer number of purchase data PCD for "NKU_003," and recommended purchase quantity information SPNJ corresponding to the recommended purchase quantity data SPND for "NKU_003." The operation screen SGC also includes a selection box CSB corresponding to a GUI for selecting customers predicted to purchase the new product. The operation screen SGD also includes the selection box NSB, the attribute information NZJ, predicted customer number of purchase information PCJ, recommended purchase quantity information SPNJ, the selection box CSB, and a pull-down list CPD. Also, the pull-down list CPD displays items corresponding to each customer included in the customer list data CLD, such as "customer001" and "customer003".

[0057] The user can understand the attributes of the new product corresponding to "NKU_003" by checking the attribute information NZJ displayed after selecting "NKU_003" in the pull-down list NPD of the operation screen SGB. The user can also understand the number of customers likely to purchase the new product corresponding to "NKU_003" by checking the predicted number of purchasers information PCJ displayed after selecting "NKU_003" in the pull-down list NPD of the operation screen SGB. The user can also consider the number of new products to purchase corresponding to "NKU_003" by referring to the recommended purchase quantity information SPNJ displayed after selecting "NKU_003" in the pull-down list NPD of the operation screen SGB. The user can also display the pull-down list CPD of the operation screen SGD by, for example, pressing the selection box CSB of the operation screen SGC. The user can also select one item from each item displayed in the pull-down list CPD.

[0058] The terminal device 200 outputs an instruction signal SS indicating that one item in the pull-down list CPD has been selected to the server device 100. Specifically, the terminal device 200 outputs an instruction signal SS indicating that, for example, "customer001" has been selected in the pull-down list CPD to the server device 100.

[0059] After outputting the display data HD corresponding to "NKU_003," if an instruction signal SS indicating that "customer001" has been selected is input from the terminal device 200, the display control unit 16D outputs the example data JGDA to the terminal device 200 as display data HD. According to this processing, for example, the operation screen SGE of FIG. 13 can be displayed on the terminal device 200. FIG. 13 is a diagram showing another example of an operation screen displayed in response to processing by the server device according to the present disclosure.

[0060] The operation screen SGE includes a selection box NSB with "NKU_003" selected and a selection box CSB with "customer001" selected. The operation screen SGE also includes attribute information NZJ, predicted customer number of purchase information PCJ, recommended stock quantity information SPNJ, and example data JGDA. For ease of illustration, the operation screen SGE in Figure 13 omits the character strings written in the nodes of the example data JGDA in Figure 8.

[0061] By checking the example data JGDA that is displayed after selecting "customer001" in the pull-down list CPD on the operation screen SGD, the user can understand the basis for the prediction result that the customer corresponding to "customer001" will purchase the new product corresponding to "NKU_003."

[0062] [Processing Flow] Next, a description will be given of the flow of processing performed in the server device 100. Fig. 14 is a flowchart showing an example of processing performed in the server device according to the present disclosure.

[0063] First, the server device 100 learns a learning model using three graph data stored in the graph data storage unit 11 and relationship data indicating known relationships between nodes linked in the three graph data (step S11).

[0064] Next, the server device 100 uses the trained learning model obtained in step S11 to predict the purchase of new products for each customer who visits a specific store (step S12).

[0065] Next, the server device 100 uses the prediction result obtained in step S12 to perform a demand forecast for the new product in the specified store (step S13). Specifically, as the processing of step S13, the server device 100 performs, for example, a process of extracting customers who are likely to purchase the new product from among all customers visiting the specified store, and a process of obtaining a recommended purchase quantity of the new product according to the number of extracted customers.

[0066] Next, the server device 100 acquires actual case data for each customer corresponding to the prediction result obtained in step S13 (step S14).

[0067] Next, the server device 100 outputs data corresponding to the instruction signal input from the terminal device 200 as display data to the terminal device 200 (step S15). Specifically, the server device 100 outputs, for example, data related to the new product processed in step S12, data related to the prediction result obtained in step S13, and data such as the example data obtained in step S14 to the terminal device 200 as display data in step S15.

[0068] As described above, according to this embodiment, for a new product for which there is no purchase history, it is possible to obtain information related to a customer group that is likely to purchase the new product and information related to the recommended purchase quantity of the new product. Also, as described above, according to this embodiment, it is possible to present to the user information related to the number of customers in a customer group that is likely to purchase the new product, information related to the recommended purchase quantity of the new product, and information indicating the basis for the prediction result that customers belonging to the customer group will purchase the new product. Therefore, according to this embodiment, it is possible to obtain information useful for demand forecasting of newly released products.

[0069] According to this embodiment, for example, the graph data EGD, CGD, and PGD for each of a plurality of stores may be stored in the graph data storage unit 11. In such a case, the purchase prediction processing unit 15 may perform a purchase prediction for customers for each of the plurality of stores, and the demand prediction processing unit 16B may perform a demand prediction for each of the plurality of stores using the prediction results of the purchase prediction.

[0070] Second Embodiment FIG. 15 is a block diagram showing an example of the functional configuration of an information processing device according to the present disclosure.

[0071] The information processing device 500 has the same hardware configuration as the server device 100. The information processing device 500 also has a learning unit 511, a purchase prediction unit 512, a demand prediction unit 513, and a display data output unit 514.

[0072] The learning means 511 can be realized, for example, by using the function of the learning processing unit 12. The purchase prediction means 512 can be realized, for example, by using the function of the purchase prediction processing unit 15. The demand prediction means 513 can be realized, for example, by using the function of the demand prediction processing unit 16B. The display data output means 514 can be realized, for example, by using the function of the display control unit 16D.

[0073] FIG. 16 is a flowchart illustrating processing performed in an information processing device according to the present disclosure.

[0074] The learning means 511 learns the learning model 600 using first graph data indicating the attributes of conventional products, second graph data indicating the attributes of customers visiting a specified store, third graph data indicating the purchasing status of the conventional products, and relationship data that is data indicating the known relationships between nodes linked in the first graph data, the second graph data, and the third graph data (step S51).

[0075] The purchase prediction means 512 uses the learned learning model 600 to predict the purchase of new products for each customer who visits a specific store (step S52).

[0076] The demand forecasting means 513 uses the results of the purchase forecast to forecast demand for new products at a predetermined store (step S53).

[0077] The display data output means 514 outputs data relating to the demand forecast results as display data (step S54).

[0078] According to this embodiment, it is possible to obtain information that is useful for predicting demand for newly released products.

[0079] A part or all of the above-described embodiments can be described as, but not limited to, the following supplementary notes.

[0080] (Supplementary Note 1) An information processing device having: a learning means for learning a learning model using first graph data indicating attributes of conventional products, second graph data indicating attributes of customers visiting a specified store, third graph data indicating purchase status of the conventional products, and relationship data which is data indicating known relationships between nodes linked in the first graph data, the second graph data, and the third graph data; a purchase prediction means for predicting purchases of new products for each customer visiting the specified store using the learned learning model; a demand prediction means for predicting demand for the new products at the specified store using prediction results of the purchase prediction; and a display data output means for outputting data related to the prediction results of the demand prediction as display data.

[0081] (Appendix 2) The information processing device described in Appendix 1, wherein the display data output means outputs as the display data a customer list indicating customers belonging to a customer group predicted to purchase the new product, a predicted number of purchasing customers indicating the number of customers belonging to the customer group, and a recommended purchase quantity of the new product obtained according to the number of customers belonging to the customer group.

[0082] (Supplementary Note 3) The information processing device described in Supplementary Note 2, wherein the purchase prediction means obtains a score according to the probability of each customer visiting the specified store purchasing the new product as a prediction result of the purchase prediction, and the demand prediction means extracts, from each customer who has a score equal to or higher than a specified score in the prediction result of the purchase prediction, customers who correspond to extraction conditions set based on the situation specific to the specified store and the attributes of the new product as customers belonging to the customer group.

[0083] (Appendix 4) The information processing device according to Appendix 3, wherein the demand forecasting means sets the number of customers corresponding to the upper limit of the number of products having the specified attribute that can be purchased at the specified store as the extraction condition when the new product has the specified attribute.

[0084] (Appendix 5) The information processing device described in Appendix 3, wherein the demand forecasting means calculates an upper limit for the purchase quantity by dividing the budget that can be allocated to the new product at the specified store by the sales price of the new product, and sets the number of customers corresponding to the calculated upper limit as the extraction condition.

[0085] (Supplementary Note 6) The information processing device according to Supplementary Note 2, wherein the demand forecasting means acquires the number of customers belonging to the customer group as the recommended purchase quantity of the new product.

[0086] (Appendix 7) The information processing device described in Appendix 2, wherein the demand forecasting means, when the new product has a specified attribute, obtains as the recommended purchase quantity of the new product a value obtained by multiplying the number of customers belonging to the customer group by a correction value for the purchase quantity applied to products having the specified attribute.

[0087] (Appendix 8) The information processing device according to Appendix 2, wherein the display data output means outputs, when a customer is selected from among the customers included in the customer list, graph data showing the basis for the prediction result of the purchase prediction corresponding to the selected customer as the display data.

[0088] (Supplementary Note 9) An information processing method executed by a computer, comprising: training a learning model using first graph data indicating attributes of conventional products, second graph data indicating attributes of customers visiting a specified store, third graph data indicating purchase status of the conventional products, and relationship data which is data indicating known relationships between nodes linked in the first graph data, the second graph data, and the third graph data; using the trained learning model, making a purchase prediction of a new product for each customer visiting the specified store; using the prediction results of the purchase prediction, making a demand prediction for the new product at the specified store; and outputting data related to the prediction results of the demand prediction as display data.

[0089] (Appendix 10) An information processing method as described in Appendix 9, which outputs as the display data a customer list indicating customers belonging to a customer group predicted to purchase the new product, a predicted number of purchasing customers indicating the number of customers belonging to the customer group, and a recommended purchase quantity of the new product obtained according to the number of customers belonging to the customer group.

[0090] (Appendix 11) An information processing method as described in Appendix 10, which obtains a score according to the probability of each customer visiting the specified store purchasing the new product as a prediction result of the purchase prediction, and extracts, from each customer who has a score equal to or higher than a predetermined score in the prediction result of the purchase prediction, customers who correspond to extraction conditions set based on the situation specific to the specified store and the attributes of the new product as customers belonging to the customer group.

[0091] (Appendix 12) An information processing method according to Appendix 11, in which, if the new product has a predetermined attribute, the number of customers corresponding to the upper limit of the number of products having the predetermined attribute that can be purchased at the predetermined store is set as the extraction condition.

[0092] (Appendix 13) An information processing method as described in Appendix 11, which calculates an upper limit on the number of purchases by dividing the budget that can be allocated to the new product at the specified store by the sales price of the new product, and sets the number of customers corresponding to the calculated upper limit as the extraction condition.

[0093] (Supplementary Note 14) The information processing method according to Supplementary Note 10, wherein the number of customers belonging to the customer group is acquired as the recommended purchase quantity of the new product.

[0094] (Supplementary Note 15) A recording medium having recorded thereon a program that causes a computer to execute the following processes: training a learning model using first graph data indicating the attributes of conventional products, second graph data indicating the attributes of customers visiting a specified store, third graph data indicating the purchasing status of the conventional products, and relationship data that is data indicating known relationships between nodes linked in the first graph data, the second graph data, and the third graph data; using the learned learning model to predict purchases of new products for each customer visiting the specified store; using the prediction results of the purchase predictions to predict demand for the new products at the specified store; and outputting data related to the prediction results of the demand forecast as display data.

[0095] (Appendix 16) A recording medium described in Appendix 15, which stores a program that causes a computer to execute a process of outputting as the display data a customer list indicating customers belonging to a customer group predicted to purchase the new product, a predicted number of purchasing customers indicating the number of customers belonging to the customer group, and a recommended purchase quantity of the new product obtained according to the number of customers belonging to the customer group.

[0096] (Appendix 17) A recording medium as described in Appendix 16, having recorded thereon a program that causes a computer to execute a process of obtaining a score according to the probability of each customer visiting the specified store purchasing the new product as a prediction result of the purchase prediction, and extracting, from each customer who has a score equal to or higher than a predetermined score in the prediction result of the purchase prediction, customers who correspond to extraction conditions set based on the situation specific to the specified store and the attributes of the new product as customers belonging to the customer group.

[0097] (Appendix 18) A recording medium as described in Appendix 17, which stores a program that causes a computer to execute a process of setting the number of customers corresponding to the upper limit of the number of products having a specified attribute that can be purchased at the specified store as the extraction condition when the new product has the specified attribute.

[0098] (Appendix 19) A recording medium described in Appendix 17 that stores a program that causes a computer to execute a process of calculating an upper limit on the number of purchases by dividing the budget that can be allocated to the new product at the specified store by the sales price of the new product, and setting the number of customers corresponding to the calculated upper limit as the extraction condition.

[0099] (Supplementary Note 20) A recording medium according to Supplementary Note 16, having recorded thereon a program for causing a computer to execute a process of acquiring the number of customers belonging to the customer group as the recommended purchase quantity of the new product.

[0100] Although the present disclosure has been described above with reference to the embodiments, the present disclosure is not limited to the above-described embodiments and examples. Various modifications that would be understood by those skilled in the art can be made to the configuration and details of the present disclosure within the scope of the present disclosure. Each embodiment can be combined with other embodiments as appropriate. Furthermore, some or all of the configurations described in Supplements 2 to 8, which are dependent on Supplement 1, may also be dependent on Supplements 9 and 15 in the same manner as Supplements 2 to 8. Furthermore, without departing from the scope of each of the above-described embodiments, not limited to Supplement 1, Supplement 9, and Supplement 15, some or all of the configurations described as Supplements may also be dependent on various hardware, software, various recording means for recording software, or systems.

Claims

1. Learning means for performing learning of a learning model using first graph data indicating the attributes of a conventional product, second graph data indicating the attributes of customers who visit a predetermined store, third graph data indicating the purchase status of the conventional product, and relational data which is data indicating known relationships between nodes linked in the first graph data, the second graph data, and the third graph data; purchase prediction means for performing a purchase prediction of a new product for each customer who visits the predetermined store using the learning model that has been learned; demand prediction means for performing a demand prediction of the new product in the predetermined store using the prediction result of the purchase prediction; and display data output means for outputting data related to the prediction result of the demand prediction as display data. An information processing apparatus having these components.

2. The information processing apparatus according to claim 1, wherein the display data output means outputs, as the display data, a customer list indicating customers belonging to a customer group predicted to purchase the new product, a predicted number of purchasing customers indicating the number of customers belonging to the customer group, and a recommended order quantity of the new product obtained according to the number of customers belonging to the customer group.

3. The purchase prediction means obtains, as a prediction result of the purchase prediction, a score corresponding to the purchase probability of the new product for each customer who visits the predetermined store. The demand prediction means extracts, as customers belonging to the customer group, customers corresponding to extraction conditions set based on the situation specific to the predetermined store and the attributes of the new product from among the customers having a score equal to or higher than a predetermined score in the prediction result of the purchase prediction. The information processing apparatus according to claim 2.

4. The information processing apparatus according to claim 3, wherein when the new product has a predetermined attribute, the demand prediction means sets, as the extraction condition, the number of customers corresponding to the upper limit value of the order quantity of the product having the predetermined attribute in the predetermined store.

5. The information processing apparatus according to claim 3, wherein the demand prediction means calculates an upper limit value of the order quantity by dividing the budget allocable to the new product in the predetermined store by the selling price of the new product, and sets, as the extraction condition, the number of customers corresponding to the calculated upper limit value.

6. The information processing apparatus according to claim 2, wherein the demand prediction means obtains the number of customers belonging to the customer group as the recommended order quantity of the new product.

7. The information processing apparatus according to claim 2, wherein the demand prediction means obtains, as the recommended purchase quantity of the new product, a value obtained by multiplying the number of customers belonging to the customer group by a correction value of the number of purchases applied to the product having the predetermined attribute when the new product has the predetermined attribute.

8. The information processing apparatus according to claim 2, wherein when one customer is selected from among the customers included in the customer list, the display data output means outputs, as the display data, graph data indicating the basis of the prediction result of the purchase prediction corresponding to the one customer.

9. An information processing method executed by a computer, the method comprising: learning a learning model using first graph data indicating attributes of a conventional product, second graph data indicating attributes of customers visiting a predetermined store, third graph data indicating a purchase situation of the conventional product, and relationship data indicating known relationships between nodes linked in the first graph data, the second graph data, and the third graph data; performing a purchase prediction of a new product for each customer visiting the predetermined store using the learned learning model; performing a demand prediction of the new product in the predetermined store using the prediction result of the purchase prediction; and outputting data related to the prediction result of the demand prediction as display data.

10. The information processing method according to claim 9, wherein a customer list indicating customers belonging to a customer group predicted to purchase the new product, a predicted number of purchasing customers indicating the number of customers belonging to the customer group, and a recommended purchase quantity of the new product obtained according to the number of customers belonging to the customer group are output as the display data.

11. The information processing method according to claim 10, wherein as a prediction result of the purchase prediction, a score corresponding to the purchase probability of the new product for each customer visiting the predetermined store is obtained, and customers corresponding to extraction conditions set based on a situation specific to the predetermined store and attributes of the new product are extracted as customers belonging to the customer group from among the customers having a score equal to or higher than a predetermined score in the prediction result of the purchase prediction.

12. The information processing method according to claim 11, wherein when the new product has a predetermined attribute, the number of customers corresponding to an upper limit value of the purchase quantity of the product having the predetermined attribute in the predetermined store is set as the extraction condition.

13. The information processing method according to claim 11, wherein an upper limit value of the purchase quantity is calculated by dividing the budget allocable to the new product in the predetermined store by the selling price of the new product, and the number of customers corresponding to the calculated upper limit value is set as the extraction condition.

14. The information processing method according to claim 10, wherein the number of customers belonging to the customer group is obtained as the recommended purchase quantity of the new product.

15. A recording medium recording a program for causing a computer to execute a process of: learning a learning model using first graph data indicating attributes of a conventional product, second graph data indicating attributes of customers visiting a predetermined store, third graph data indicating a purchase status of the conventional product, and relationship data which is data indicating known relationships between nodes linked in the first graph data, the second graph data, and the third graph data; predicting a purchase of a new product for each customer visiting the predetermined store using the learned learning model; predicting a demand for the new product in the predetermined store using the prediction result of the purchase prediction; and outputting data related to the prediction result of the demand prediction as display data.

16. The recording medium according to claim 15, recording a program for causing a computer to execute a process of outputting, as the display data, a customer list indicating customers belonging to a customer group predicted to purchase the new product, a predicted number of purchasing customers indicating the number of customers belonging to the customer group, and a recommended purchase quantity of the new product obtained according to the number of customers belonging to the customer group.

17. The recording medium according to claim 16, recording a program for causing a computer to execute a process of: obtaining, as a prediction result of the purchase prediction, a score corresponding to the purchase probability of the new product for each customer visiting the predetermined store; and extracting, as customers belonging to the customer group, customers corresponding to extraction conditions set based on the situation specific to the predetermined store and the attributes of the new product from among the customers having a score equal to or higher than a predetermined score in the prediction result of the purchase prediction.

18. The recording medium according to claim 17, recording a program for causing a computer to execute a process of setting, as the extraction condition, the number of customers corresponding to the upper limit value of the purchase quantity of the product having the predetermined attribute in the predetermined store when the new product has the predetermined attribute.

19. The recording medium according to claim 17, which records a program for causing a computer to execute a process of calculating an upper limit value of the purchase quantity by dividing the budget assignable to the new product in the predetermined store by the selling price of the new product, and setting the number of customers corresponding to the calculated upper limit value as the extraction condition.

20. The recording medium according to claim 16, which records a program for causing a computer to execute a process of obtaining the number of customers belonging to the customer group as the recommended purchase quantity of the new product.

Images