Prediction device, prediction method, and program

Abstract

In this prediction device, a graph data generation means generates graph data including a plurality of nodes and a link indicating the relationship between the nodes from commodity information, customer information, and a purchase history. A new commodity generation means extracts commodity features from the graph data, and generates a new commodity composed of a new commodity feature combination. A new commodity addition means adds the new commodity to the graph data. A model generation means generates, by using graph data to which the new commodity is added, a prediction model for predicting a purchase probability from combinations between commodities and customers. A customer feature generation means extracts customer features from the graph data to which the new commodity is added, and generates a combination of the customer features. An acquisition means acquires the new commodity and the combination of the customer features. A prediction means predicts, by using the prediction model, a purchase probability at which a customer having the combination of the customer features purchases the new commodity.

Description

Prediction device, prediction method, and recording medium

[0001] The present disclosure relates to prediction techniques using graph data.

[0002] In recent years, systems that support new product planning and development work have become known, but the creation of concepts in new product planning often depends on the experience and knowledge of the person in charge. Therefore, it is desirable to be able to plan new products based on past data, etc., without relying on the experience and knowledge of the person in charge. For example, Patent Document 1 proposes a design support device that can evaluate the impression of a product by combining design elements related to the parts of the product and support the improvement of the impression.

[0003] International Publication No. WO2019-028148

[0004] The method of Patent Document 1 requires evaluation values ​​for each product feature for each target segment as training data, but in order to obtain highly reliable evaluation values, a large number of monitor surveys are required, which is time-consuming and labor-intensive.

[0005] One object of the present disclosure is to provide a prediction device that can predict new products that will have a high purchase probability by using existing purchase history, etc.

[0006] In one aspect of the present disclosure, a prediction device comprises: a graph data generation means for generating graph data including a plurality of nodes and links indicating relationships between the nodes from product information, customer information, and purchase history; a new product generation means for extracting product features from the graph data and generating a new product consisting of a new combination of product features; a new product addition means for adding the new product to the graph data; a model generation means for generating a prediction model for predicting a purchase probability from a combination of product and customer, using the graph data to which the new product has been added; a customer feature generation means for extracting customer features from the graph data to which the new product has been added, and generating a combination of customer features; an acquisition means for acquiring the new product and the combination of customer features; and a prediction means for predicting a purchase probability that a customer having the combination of customer features will purchase the new product, using the prediction model.

[0007] In yet another aspect of the present disclosure, a prediction method includes generating graph data including multiple nodes and links indicating relationships between the nodes from product information, customer information, and purchase history; extracting product features from the graph data and generating new products consisting of new combinations of product features; adding the new products to the graph data; using the graph data with the new products added, generating a prediction model that predicts a purchase probability from a combination of product and customer; extracting customer features from the graph data with the new products added and generating a combination of customer features; obtaining the new product and the combination of customer features; and using the prediction model, predicting the purchase probability that a customer having the combination of customer features will purchase the new product.

[0008] In yet another aspect of the present disclosure, a recording medium records a program that causes a computer to execute the following processes: generate graph data including a plurality of nodes and links indicating relationships between the nodes from product information, customer information, and purchase history; extract product features from the graph data, and generate new products consisting of new combinations of product features; add the new products to the graph data; use the graph data with the new products added to generate a prediction model that predicts purchase probability from combinations of products and customers; extract customer features from the graph data with the new products added, and generate combinations of customer features; obtain the new products and the combinations of customer features; and use the prediction model to predict the purchase probability that customers having the combination of customer features will purchase the new product.

[0009] According to the present disclosure, it is possible to predict new products that will have a high purchase probability by using existing purchase history, etc.

[0010] 1 shows the overall configuration of a prediction device according to a first embodiment; FIG. 2 is a block diagram showing the hardware configuration of the prediction device according to the first embodiment; FIG. 3 is a diagram showing the functional configuration of the prediction device according to the first embodiment; FIG. 4 shows an example of the data structure of customer information, product information, and purchase history; FIG. 5 is a schematic diagram showing graph data; FIG. 6 is a schematic diagram showing graph data to which a new product has been added; FIG. 7 is a flowchart showing processing by a learning unit of the prediction device according to the first embodiment; FIG. 8 shows an example of an output result by the prediction device according to the first embodiment; FIG. 9 is a flowchart showing processing by a prediction unit of the prediction device according to the first embodiment;

[0011] Hereinafter, preferred embodiments of the present disclosure will be described with reference to the drawings.

[0012] [System Configuration] Fig. 1 shows a prediction device according to this embodiment. The prediction device 10 learns a prediction model based on input data such as product information, customer information, and purchase history. The prediction device 10 then uses the prediction model to predict the probability that a customer will purchase a new product.

[0013] 2 is a block diagram showing the hardware configuration of the prediction device 10. As shown in the figure, the prediction device 10 includes an interface (IF) 11, a processor 12, a memory 13, a recording medium 14, a database (DB) 15, an input unit 16, and a display unit 17.

[0014] The IF 11 inputs and outputs data to and from an external device. Specifically, the prediction device 10 acquires product information, customer information, and purchase history through the IF 11. Furthermore, the prediction results of the prediction device 10 are output to the external device through the IF 11 as needed.

[0015] The processor 12 is a computer such as a CPU (Central Processing Unit) and controls the entire prediction device 10 by executing a pre-prepared program. The processor 12 may be a CPU, a GPU (Graphics Processing Unit), a DSP (Digital Signal Processor), an MPU (Micro Processing Unit), an FPU (Floating Point number Processing Unit), a PPU (Physics Processing Unit), a TPU (Tensor Processing Unit), a quantum processor, a microcontroller, or a combination thereof. Specifically, the processor 12 executes a prediction model generation process and a prediction process, which will be described later.

[0016] The memory 13 is configured by a ROM (Read Only Memory), a RAM (Random Access Memory), etc. The memory 13 is also used as a working memory while the processor 12 is executing various processes.

[0017] The recording medium 14 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 prediction device 10. The recording medium 14 records various programs to be executed by the processor 12. When the prediction device 10 executes various processes, the programs recorded on the recording medium 14 are loaded into the memory 13 and executed by the processor 12.

[0018] The DB 15 stores product information, customer information, purchase history, and the like input via the IF 11. The DB 15 also stores graph data generated by the prediction device 10 as a graph database.

[0019] The input unit 16 is, for example, a keyboard, a mouse, etc., and is used by the user to input input data that will be the basis for prediction.

[0020] The display unit 17 is, for example, a liquid crystal display device, and displays the prediction results obtained by the prediction device 10 .

[0021] 3 is a block diagram showing the functional configuration of the prediction device 10. Functionally, the prediction device 10 includes a learning unit 110 and a prediction unit 120 in addition to the DB 15 described above.

[0022] (Learning Unit) First, we will explain the learning unit 110. The learning unit 110 includes a graph conversion unit 111, a product feature extraction unit 112, a product feature combination unit 113, a new product data addition unit 114, a prediction model learning unit 115, and a prediction model output unit 116.

[0023] Product information, customer information, and purchase history are input to the prediction device 10 via the IF 11. The graph conversion unit 111 generates graph data based on the product information, customer information, and purchase history, and outputs the graph data to the DB 15.

[0024] FIG. 4 shows examples of product information, customer information, and purchase history. FIG. 4(A) shows an example of product information. Product information includes a product ID that uniquely identifies the product, a product name, product characteristics, etc. Product characteristics indicate the characteristics of the product, such as sweet, bitter, high calorie, low calorie, or sales replenishment. FIG. 4(B) shows an example of customer information. Customer information includes a customer ID that uniquely identifies the customer, the customer's name, customer characteristics, etc. Customer characteristics indicate the customer's attributes, such as the customer's age, gender, and address. FIG. 4(C) shows an example of purchase history. The purchase history includes a purchase number that uniquely identifies the transaction, the customer's name, and the product name. Note that the product information, customer information, and purchase history shown in FIG. 4 are merely examples and are not limited to these.

[0025] FIG. 5 schematically illustrates graph data generated by the graph conversion unit 111. Graph data 50 in FIG. 5 represents product information, customer information, and purchase history in a knowledge graph. The graph data includes nodes, links, and relations. Nodes correspond to product names, product features, customer names, and customer features. Links indicate connections between nodes. Relations indicate relationships between nodes. For example, in the graph data 50, a link is added between the node "Customer A" and the node "Female," indicating that the relationship is "Gender." Furthermore, in the graph data 50, a link is added between the node "Product Y" and the node "Sweet," indicating that the relationship is "Characteristics." Furthermore, in the graph data 50, a link is added between the node "Customer A" and the node "Product Y," indicating that the relationship is "Purchase."

[0026] 3 , the product feature extraction unit 112 acquires graph data from the DB 15. Then, the product feature extraction unit 112 extracts product features from the graph data and generates a product feature list. The product feature extraction unit 112 outputs the product feature list to the product feature combination unit 113.

[0027] The product feature combination unit 113 generates a new product based on the product feature list input from the product feature extraction unit 112. Specifically, the product feature combination unit 113 selects one or more product features from the product feature list and generates one or more new combinations of product features. A product consisting of a new combination of product features will hereinafter be referred to as a "new product." The product feature combination unit 113 outputs the new product to the new product data addition unit 114.

[0028] The new product data adding unit 114 adds the new products input from the product feature combination unit 113 to the graph data in the DB 15. Fig. 6 schematically shows the graph data to which the new products have been added. The graph data 60 in Fig. 6 is obtained by adding new product Z to the graph data 50 in Fig. 5. New product Z has the product features of "low calorie," "nutritious," and "bitter," and links are provided between new product Z and each of the product features of new product Z.

[0029] The prediction model training unit 115 acquires graph data from DB 15. Then, the prediction model training unit 115 uses the graph data to learn the relationships between new products and customers and generate a prediction model. The prediction model training unit 115 trains the prediction model by inputting, for example, feature values ​​corresponding to each node included in the graph data and feature values ​​corresponding to each relationship based on the KBLRN. FIG. 7 schematically illustrates the trained graph data. For the graph data 70 in FIG. 7 , the prediction model training unit 115 performs training so that relationships between unlinked nodes can be derived from known relationships between linked nodes in the graph data. For the graph data 70, the prediction model training unit 115 learns whether a link should be added between new product Z and a customer. The prediction model training unit 115 outputs the generated prediction model to the prediction model output unit 116.

[0030] KBLRN is a framework for graph-based relational learning, and the method described in the following document can be used: Alberto Garcia-Duran and Mathias Niepert: KBLRN: End-to-End Learning of Knowledge Base Representations with Latent, Relational, and Numerical Features, the 34th Conference on Uncertainty in Artificial Intelligence (UAI)

[0031] The prediction model output unit 116 outputs the prediction model input from the prediction model learning unit 115 to the purchase score calculation unit 123 described below.

[0032] (Prediction Model Generation Process) Next, the prediction model generation process will be described. Fig. 8 is a flowchart of the prediction model generation process performed by the learning unit 110. This process is realized by the processor 12 shown in Fig. 2 executing a prepared program and operating as each element shown in Fig. 3.

[0033] First, product information, customer information, and purchase history are input to the prediction device 10 via the IF 11. The graph conversion unit 111 generates graph data based on the product information, customer information, and purchase history, and outputs the graph data to the DB 15 (steps S111 and S112).

[0034] Next, the product feature extraction unit 112 acquires graph data from DB15. Then, the product feature extraction unit 112 extracts product features from the graph data and generates a product feature list. The product feature extraction unit 112 outputs the product feature list to the product feature combination unit 113 (step S113). Next, the product feature combination unit 113 generates new products based on the product feature list input from the product feature extraction unit 112. Specifically, the product feature combination unit 113 selects one or more product features from the product feature list and generates one or more new combinations of product features. The product feature combination unit 113 outputs the new products to the new product data addition unit 114 (step S114). Next, the new product data addition unit 114 adds the new products input from the product feature combination unit 113 to the graph data in DB15 (step S115).

[0035] Next, the prediction model training unit 115 acquires the graph data from the DB 15. Then, the prediction model training unit 115 uses the graph data to learn the relationship between the new product and the customer, and generates a prediction model. The prediction model training unit 115 outputs the generated prediction model to the prediction model output unit 116 (step S116).

[0036] Next, the prediction model output unit 116 outputs the prediction model input from the prediction model learning unit 115 to the purchase score calculation unit 123 (step S117), which will be described later, and the prediction model generation process ends.

[0037] (Prediction Unit) Next, a description will be given of the prediction unit 120. The prediction unit 120 includes a customer feature extraction unit 121, a customer feature combination unit 122, a purchase score calculation unit 123, and a purchase score output unit .

[0038] The customer feature extraction unit 121 acquires graph data from the DB 15. Then, the customer feature extraction unit 121 extracts customer features from the graph data and generates a customer feature list. The customer feature extraction unit 121 outputs the customer feature list to the customer feature combination unit 122.

[0039] The customer feature combination unit 122 selects one or more customer features from the customer feature list input from the customer feature extraction unit 121 and generates one or more combinations of customer features. A customer having the generated combination of customer features will hereinafter also be referred to as a "target customer." The target customer is used as input data when predicting the probability of purchasing a new product. The customer feature combination unit 122 outputs the target customer to the purchase score calculation unit 123.

[0040] The purchase score calculation unit 123 acquires graph data from DB 15. The purchase score calculation unit 123 also receives the prediction model from the prediction model output unit 116 and the target customers from the customer feature combination unit 122. The purchase score calculation unit 123 uses the prediction model to predict the purchase probability of each new product for each target customer. The purchase score calculation unit 123 then outputs the prediction result to the purchase score output unit 124.

[0041] The purchase score calculation unit 123 may predict the purchase probability for all new products, i.e., for all new combinations of product features included in the graph data, or may predict the purchase probability for only predetermined new products. The predetermined new products can be specified, for example, by a user via the input unit 16 of the prediction device 10. The purchase score calculation unit 123 may predict the purchase probability for all target customers, i.e., for all combinations of customer features included in the graph data, or may predict the purchase probability for customers having a predetermined combination of customer features. The predetermined combination of customer features can be specified, for example, by a user via the input unit 16 of the prediction device 10.

[0042] FIG. 9 shows an example of a prediction result. FIG. 9(A) shows the purchase probability of each new product for each target customer. For example, in FIG. 9(A), the prediction device 10 predicts that the purchase probability of new product 1 for customer A is 50%. FIG. 9(B) shows the purchase probability of each new product for each set of target customers. A set of target customers is a set of customers having a predetermined combination of customer characteristics. The prediction device 10 predicts the purchase probability of each customer included in the set of target customers and calculates a representative value such as the average or mode, thereby outputting the prediction result. For example, in FIG. 9(B), the prediction device 10 predicts that the purchase probability of new product 1 is 50% for a set of customers having the combination of "in their 30s" and "female" as the predetermined customer characteristics.

[0043] The purchase score calculation unit 123 may output the optimal combination of the target customer and the new product or the optimal new product as a prediction result to the purchase score output unit 124. For example, the purchase score calculation unit 123 may output, as the optimal combination, a combination of the target customer and the new product that results in the highest purchase probability among the purchase probabilities of each new product for each group of target customers to the purchase score output unit 124. Furthermore, the purchase score calculation unit 123 may output, as the optimal new product, a new product that results in the highest purchase probability among the purchase probabilities of each new product for each group of target customers. Furthermore, the purchase score calculation unit 123 may calculate the average purchase probability of the target customers for each new product, and output the new product with the highest average purchase probability to the purchase score output unit 124 as the optimal new product.

[0044] (Prediction Processing) Next, the prediction processing will be described. Figures 10 and 11 are flowcharts of the prediction processing by the prediction unit 120. This processing is realized by the processor 12 shown in Figure 2 executing a program prepared in advance and operating as each element shown in Figure 3.

[0045] In the prediction process of Figure 10, first, the customer feature extraction unit 121 acquires graph data from DB 15 (step S121). Then, the customer feature extraction unit 121 extracts customer features from the graph data and generates a customer feature list (step S122). Next, the customer feature combination unit 122 generates target customers from the customer feature list input from the customer feature extraction unit 121. Specifically, the customer feature combination unit 122 selects one or more customer features from the customer feature list and generates one or more target customers. The customer feature combination unit 122 outputs the target customers to the purchase score calculation unit 123 (step S123).

[0046] The purchase score calculation unit 123 receives the prediction model from the prediction model output unit 116 (step S124). The purchase score calculation unit 123 uses the prediction model to predict the purchase probability of each new product for each target customer. The purchase score calculation unit 123 then outputs the prediction result to the purchase score output unit 124 (step S125).

[0047] The purchase score output unit 124 outputs the prediction result to the DB 15, the display unit 17, etc. (step S126), and the process ends.

[0048] Next, the prediction process of Fig. 11 will be described. Fig. 11 differs from the flowchart of the prediction process of Fig. 10 in that the user specifies the new product and customer to be predicted. Note that the processes of steps S131 to S134 of Fig. 11 are the same as the processes of steps S121 to S124 of Fig. 10, and therefore their description will be omitted.

[0049] The user specifies a new product and a combination of customer features via the input unit 16 of the prediction device 10. The purchase score calculation unit 123 acquires the new product and combination of customer features specified by the user (step S135). The purchase score calculation unit 123 uses a prediction model to predict the purchase probability of the specified new product for each customer having the specified combination of customer features. The purchase score calculation unit 123 then outputs the prediction result to the purchase score output unit 124 (step S136). The purchase score output unit 124 outputs the prediction result to the DB 15, the display unit 17, etc. (step S137), and the processing ends.

[0050] In the above configuration, the graph conversion unit 111 is an example of a graph data generation means, the product feature extraction unit 112 and the product feature combination unit 113 are examples of a new product generation means, the new product data addition unit 114 is an example of a new product addition means, the prediction model learning unit 115 and the prediction model output unit 116 are examples of a model generation means, the customer feature extraction unit 121 and the customer feature combination unit 122 are examples of a customer feature generation means, the purchase score calculation unit 123 is an example of an acquisition means and a prediction means, and the purchase score output unit 124 is an example of a prediction means.

[0051] [Modifications] Next, a description will be given of modifications of the first embodiment. The following modifications can be applied to the first embodiment in appropriate combinations.

[0052] (Variation 1) In the first embodiment, the prediction device 10 predicts the purchase probability of a new product. However, the prediction device 10 may also predict product features to be added to the new product. For example, the prediction device 10 adds product features to a new product and predicts the purchase probability for the new product to which the product features have been added. If the purchase probability is equal to or greater than a predetermined threshold, the prediction device 10 outputs the product features added to the new product. This allows the user to understand which product features should be added to the new product to increase the purchase probability.

[0053] (Variation 2) The prediction device 10 can also be applied in the field of pharmaceuticals. For example, the prediction device 10 can predict the probability that a new pharmaceutical will be effective for a patient. The prediction device 10 learns a prediction model based on input data such as pharmaceutical information, patient information, and prescription history. In this case, the prediction device 10 uses symptoms as patient characteristics and efficacy and effects as pharmaceutical characteristics. Then, the prediction device 10 uses the prediction model to predict the probability that a new pharmaceutical will be effective for a patient.

[0054] 12 is a block diagram showing the functional configuration of a prediction device according to Embodiment 2. The prediction device 300 includes a graph data generation unit 301, a new product generation unit 302, a new product addition unit 303, a model generation unit 304, a customer feature generation unit 305, an acquisition unit 306, and a prediction unit 307.

[0055] FIG. 13 is a flowchart of processing by the prediction device of the second embodiment. The graph data generation means 301 generates graph data including multiple nodes and links indicating relationships between the nodes from product information, customer information, and purchase history (step S301). The new product generation means 302 extracts product features from the graph data and generates new products consisting of new product feature combinations (step S302). The new product addition means 303 adds the new products to the graph data (step S303). The model generation means 304 uses the graph data to which the new products have been added to generate a prediction model that predicts purchase probabilities from combinations of products and customers (step S304). The customer feature generation means 305 extracts customer features from the graph data to which the new products have been added and generates combinations of customer features (step S305). The acquisition means 306 acquires the new products and the combinations of customer features (step S306). The prediction means 307 uses the prediction model to predict the purchase probability that a customer having the combination of customer features will purchase the new product (step S307).

[0056] According to the prediction device 300 of the second embodiment, it is possible to predict new products that will have a high purchase probability by using existing purchase history and the like.

[0057] A part or all of the above-described embodiments can be described as, but not limited to, the following supplementary notes.

[0058] (Supplementary Note 1) A prediction device comprising: a graph data generation means for generating graph data including a plurality of nodes and links indicating relationships between the nodes from product information, customer information, and purchase history; a new product generation means for extracting product features from the graph data and generating a new product consisting of a new combination of product features; a new product addition means for adding the new product to the graph data; a model generation means for generating a prediction model for predicting a purchase probability from a combination of product and customer, using the graph data to which the new product has been added; a customer feature generation means for extracting customer features from the graph data to which the new product has been added, and generating a combination of customer features; an acquisition means for acquiring the new product and the combination of customer features; and a prediction means for predicting a purchase probability, using the prediction model, of a customer having the combination of customer features, that will purchase the new product.

[0059] (Supplementary Note 2) The prediction device according to Supplementary Note 1, wherein the prediction means outputs an optimal combination of a customer and a new product, or an optimal new product, based on the result of the prediction.

[0060] (Supplementary Note 3) The prediction device according to Supplementary Note 2, wherein the prediction means outputs, as an optimal combination, a combination with the highest purchase probability from among combinations of customers having the combination of customer characteristics and the new product.

[0061] (Supplementary Note 4) The prediction device according to Supplementary Note 1, wherein the graph data generation means generates graph data by linking customers and customer features with customer feature links, linking products and product features with product feature links, and linking products and customers with purchase links based on the product information, the customer information, and the purchase history.

[0062] (Supplementary Note 5) The prediction device according to Supplementary Note 1 or 2, wherein the acquisition means acquires a predetermined new product from all new products generated by the new product generation means, and the prediction means predicts a purchase probability for the predetermined new product.

[0063] (Supplementary Note 6) The prediction device according to Supplementary Note 1 or 2, wherein the acquisition means acquires a predetermined combination of customer features from all combinations of customer features generated by the customer feature generation means, and the prediction means predicts a purchase probability for a customer having the predetermined combination of customer features.

[0064] (Supplementary Note 7) A prediction device according to Supplementary Note 1, further comprising a product feature adding means for adding product features to the new product, wherein the prediction means predicts a purchase probability for the new product to which the product features have been added, and outputs the added product features if the purchase probability is equal to or greater than a predetermined threshold.

[0065] (Supplementary Note 8) A prediction method comprising: generating graph data including a plurality of nodes and links indicating relationships between the nodes from product information, customer information, and purchase history; extracting product features from the graph data and generating new products consisting of new combinations of product features; adding the new products to the graph data; using the graph data with the new products added, generating a prediction model that predicts purchase probability from combinations of products and customers; extracting customer features from the graph data with the new products added and generating combinations of customer features; obtaining the new products and the combinations of customer features; and using the prediction model to predict the purchase probability that customers having the combination of customer features will purchase the new product.

[0066] (Supplementary Note 9) A recording medium having recorded thereon a program that causes a computer to execute the following processes: generating graph data including a plurality of nodes and links indicating the relationships between the nodes from product information, customer information, and purchase history; extracting product features from the graph data and generating new products consisting of new combinations of product features; adding the new products to the graph data; using the graph data with the new products added, generating a prediction model that predicts the purchase probability from combinations of products and customers; extracting customer features from the graph data with the new products added and generating combinations of customer features; obtaining the new products and the combinations of customer features; and using the prediction model, predicting the purchase probability that customers having the combination of customer features will purchase the new product.

[0067] Although the present disclosure has been described above with reference to the embodiments and examples, the present disclosure is not limited to the above-described embodiments and examples. Various modifications that can be understood by a person skilled in the art can be made to the configuration and details of the present disclosure within the scope of the present disclosure.

[0068] REFERENCE SIGNS LIST 10 Prediction device 110 Learning unit 111 Graph conversion unit 112 Product feature extraction unit 113 Product feature combination unit 114 New product data addition unit 115 Prediction model learning unit 116 Prediction model output unit 120 Prediction unit 121 Customer feature extraction unit 122 Customer feature combination unit 123 Purchase score calculation unit 124 Purchase score output unit

Claims

1. graph data generation means for generating graph data including a plurality of nodes and links indicating relationships between the nodes from product information, customer information, and purchase history; a new product generation means for extracting product features from the graph data and generating new products each having a new combination of product features; a new product adding means for adding the new product to the graph data; a model generation means for generating a prediction model that predicts a purchase probability from a combination of a product and a customer using the graph data to which the new product has been added; a customer feature generation means for extracting customer features from the graph data to which the new product has been added and generating a combination of customer features; an acquisition means for acquiring the combination of the new product and the customer characteristics; a prediction means for predicting a purchase probability of a customer having the combination of customer characteristics purchasing the new product using the prediction model; A prediction device comprising:

2. 2. The prediction device according to claim 1, wherein the prediction means outputs an optimal combination of a customer and a new product, or an optimal new product, based on the result of the prediction.

3. The prediction device according to claim 2 , wherein the prediction means outputs, as an optimal combination, a combination with the highest purchase probability from among combinations of customers having the combination of customer characteristics and the new product.

4. 2. The prediction device according to claim 1, wherein the graph data generation means generates graph data by linking customers and customer features with customer feature links, linking products and product features with product feature links, and linking products and customers with purchase links based on the product information, the customer information, and the purchase history.

5. the acquiring means acquires a predetermined new product from all new products generated by the new product generating means; 3. The prediction device according to claim 1, wherein the prediction means predicts the purchase probability of a predetermined new product.

6. the acquisition means acquires a predetermined combination of customer features from all the combinations of customer features generated by the customer feature generation means; 3. The prediction device according to claim 1, wherein the prediction means predicts a purchase probability for a customer having a predetermined combination of customer characteristics.

7. a product feature adding means for adding a product feature to the new product; The prediction device according to claim 1 , wherein the prediction means predicts a purchase probability for the new product to which the product feature has been added, and outputs the added product feature if the purchase probability is equal to or greater than a predetermined threshold.

8. A computer-implemented prediction method comprising: Generate graph data including a plurality of nodes and links indicating relationships between the nodes from the product information, customer information, and purchase history; extracting product features from the graph data and generating new products consisting of new product feature combinations; Add the new product to the graph data, Using the graph data to which the new product has been added, a prediction model is generated that predicts the purchase probability from a combination of product and customer; extracting customer features from the graph data to which the new product has been added, and generating combinations of customer features; Acquire the combination of the new product and the customer characteristics; A prediction method that uses the prediction model to predict the purchase probability that a customer having the combination of customer characteristics will purchase the new product.

9. Generate graph data including a plurality of nodes and links indicating relationships between the nodes from the product information, customer information, and purchase history; extracting product features from the graph data and generating new products consisting of new product feature combinations; Add the new product to the graph data, Using the graph data to which the new product has been added, a prediction model is generated that predicts the purchase probability from a combination of product and customer; extracting customer features from the graph data to which the new product has been added, and generating combinations of customer features; Acquire the combination of the new product and the customer characteristics; A program that causes a computer to execute a process of predicting the purchase probability of a customer having the combination of customer characteristics purchasing the new product using the prediction model.