Recommended product determination device, recommended product determination method, and recommended product determination program

Abstract

This invention provides a product recommendation device, a product recommendation method, and a product recommendation program that enable the recommendation of products tailored to a user's preferences, taking into account when the user purchased the product. [Solution] In the communication system, the online shopping server 1, which is a recommended product determination device, acquires the user's purchase history information of products, identifies related products that have a predetermined relationship with the purchased product indicated by the purchase history information, calculates the similarity between the product relationship information of the purchased product and the product relationship information of the related product, determines recommended products from among the related products that satisfy the conditions for products to be recommended to the user based on the age of the purchase date of the purchased product and the calculated similarity, changes the degree of satisfaction of either the similarity or the age of the purchase date according to the other, and transmits recommended product information indicating the recommended product to the user terminal.

Description

Technical Field

[0001] The present invention relates to a method for determining products to be recommended to users.

Background Art

[0002] Conventionally, a method for recommending products to a user based on products the user has purchased in the past is known. For example, in Patent Document 1, a similarity between products is calculated for each pair of products, and for each product that a user logged in to a shopping site has purchased in the past, another product whose similarity is greater than a threshold value is specified as a recommended product.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, even if products similar to those the user has just purchased recently are recommended, the user may not want those products.

[0005] The present invention has been made in view of the above points, and an example of the problem is to provide a recommended product determination device, a recommended product determination method, and a recommended product determination program that enable recommending products according to the preferences of the user in consideration of the purchase timing of products by the user.

Means for Solving the Problems

[0006] One aspect of the present invention is purchase history information showing the purchase history of a target user, comprising: purchase history information acquisition means for acquiring purchase history information showing the purchased product and the time when the product was purchased; identification means for identifying related products having a predetermined relationship with the purchased product shown by the acquired purchase history information; and product relationship information acquisition means for acquiring product relationship information relating to a product, comprising: product relationship information acquisition means for acquiring the product relationship information of the purchased product and the product relationship information of the identified related product; and the product relationship information of the purchased product and the product relationship information of the identified related product. A recommended product determination device comprising: a calculation means for calculating similarity with product-related information; a determination means for determining a recommended product from among the identified related products that satisfies the conditions for a product to be recommended to the user, based on the age of the purchase date of the purchased product and the calculated similarity; and a recommended product information transmission means for transmitting recommended product information indicating the determined recommended product to the user's terminal device, wherein the conditions for the recommended product include a condition for either the similarity or the age of the purchase date, and the determination means changes the degree of ease of satisfying the condition for one of them according to the other.

[0007] According to this aspect, related products that have a predetermined relationship with the products purchased by the target user are identified. The similarity between the product relationship information of the purchased product and the product relationship information of the related products is calculated. Then, recommended products are determined based on the purchase date and similarity of the products purchased by the target user. The conditions for recommended products include conditions for either similarity or the age of purchase. When determining recommended products, the ease of satisfying one of the conditions for similarity or the age of purchase changes depending on the other. For example, if the conditions for recommended products include a condition for similarity, the ease of satisfying the condition for similarity changes depending on the age of purchase. For example, if the conditions for recommended products include a condition for the age of purchase, the ease of satisfying the condition for the age of purchase changes depending on the similarity. Therefore, by considering the timing of the user's purchase, it is possible to recommend products that match the user's preferences.

[0008] Another aspect of the present invention is a recommended product determination device characterized in that the conditions for the recommended product include the result of comparing the similarity with a reference value for the similarity to obtain a predetermined result, and the determination means changes the similarity by correcting the calculated similarity according to the age of the purchase date, or changes the reference value according to the age of the purchase date.

[0009] From this perspective, the degree to which the conditions for similarity are easily met can be changed according to the age of the purchase, either by adjusting the similarity score according to the age of the purchase, or by changing the baseline value according to the age of the purchase.

[0010] Another aspect of the present invention is a recommended product determination device characterized in that the determination means corrects the similarity or determines the reference value such that the older the purchase date, the more likely the comparison result is to be the predetermined result.

[0011] According to this aspect, the older the purchase date, the more likely the comparison results are to yield a predetermined outcome. Therefore, the older the purchase date, the easier it is to satisfy the conditions for similarity. Consequently, the older the purchase date of a product by a user, the higher the probability that the related product will be a recommended product.

[0012] A further aspect of the present invention is a recommended product determination device characterized in that the conditions for the recommended product include the similarity being less than the threshold value, and the determination means corrects the similarity so that the corrected similarity becomes lower the older the purchase date, or determines the threshold value so that the threshold value becomes higher.

[0013] According to this perspective, the older the purchase date, the easier it is to satisfy the similarity criteria, and the lower the similarity, the higher the probability of satisfying those criteria. Users purchase a product because they desire it. Once a product is purchased, the user's desire to purchase that product is satisfied. Therefore, users may not desire related products that are similar to products they have recently purchased. Related products similar to products purchased relatively recently are less likely to satisfy the similarity criteria, and thus have a relatively low probability of becoming recommended products. Since these related products are similar to the products the user desired, they may suit the user's preferences. Therefore, as time passes since the user's purchase, the user's desire to purchase related products similar to the purchased product may increase. Related products similar to products purchased relatively long ago are more likely to satisfy the similarity criteria than those similar to products purchased relatively recently, and thus have a relatively higher probability of becoming recommended products. Even if a related product is not similar to the product a user has purchased, that related product still has a certain relationship to the product the user purchased. Therefore, that related product may suit the user's preferences. Even if the product a user purchased was relatively recent, a related product that is not similar to that product is more likely to be recommended than a related product that is similar. Therefore, by considering when a user purchased a product, it is possible to recommend products that match the user's preferences.

[0014] Another aspect of the present invention is a recommended product determination device characterized in that the determination means determines a correction coefficient for the similarity according to the age of the purchase date, and calculates the corrected similarity based on the determined correction coefficient and the similarity.

[0015] From this perspective, by determining a correction factor for similarity based on the age of the purchase, the degree to which the conditions for similarity are easily satisfied can be changed.

[0016] Another aspect of the present invention is a recommended product determination device characterized in that the conditions for the recommended product include conditions for similarity, and the determination means increases the degree to which the conditions for similarity are easily satisfied as the purchase date is older.

[0017] According to this perspective, the older a product is purchased by a user, the higher the probability that related products will become recommended products.

[0018] A further aspect of the present invention is a recommended product determination device characterized in that the purchased goods include a first product and a second product, the calculation means calculates a first similarity between the product relationship information of the related product and the product relationship information of the first product, and if the determination means determines that the related product satisfies the conditions for the recommended product based on the first similarity and the age of the purchase date of the first product, it calculates a second similarity between the product relationship information of the related product and the product relationship information of the second product, and if the determination means determines that the related product does not satisfy the conditions for the recommended product based on the first similarity and the age of the purchase date of the first product, it does not calculate the second similarity.

[0019] According to this aspect, if a user has purchased multiple products, a first similarity score is calculated for any first product between it and the related product. Based on the first similarity score and the age of the first product's purchase, it is determined whether the related product meets the criteria for a recommended product. If the related product meets the criteria, a second similarity score is calculated for any second product among the purchased products between it and the related product. If the related product does not meet the criteria, the second similarity score is not calculated. Therefore, when calculating similarity between each of the related products and the purchased products to determine a recommended product, unnecessary similarity calculations can be omitted.

[0020] Another aspect of the present invention is a recommended product determination device characterized in that the determination means determines the related products that are determined to satisfy the conditions for recommended products for all of the purchased products as the recommended products.

[0021] According to this aspect, for any of the products purchased by the user, it is possible to determine a product suitable as a recommended product.

[0022] Still another aspect of the present application is that the acquired product-related information includes a vector of text and a vector of an image of the product posted on the website of a trading service that enables the purchase of the product, for the product to which the product-related information corresponds, and the calculating means calculates the similarity based on the vector of the text and the vector of the image. The recommended product determination device is characterized by this.

[0023] According to this aspect, the similarity is calculated based on the vectors of text and the vectors of images for each of the purchased product and related products. Therefore, the similarity between the purchased product and the related products can be more appropriately determined.

[0024] Still another aspect of the present application is that the acquired product-related information includes a vector indicating the browsing status of whether a plurality of products on the website of a trading service that enables the purchase of the product and the product to which the product-related information corresponds have been browsed by the same user at the same time, and a vector of text posted on the website for the product to which the product-related information corresponds, and the calculating means calculates the similarity based on the vector indicating the browsing status and the vector of the text. The recommended product determination device is characterized by this.

[0025] According to this aspect, the similarity is calculated based on the vectors indicating the browsing status and the vectors of text for each of the purchased product and related products with other products at the same time. Therefore, the similarity between the purchased product and the related products can be more appropriately determined.

[0026] Another aspect of the present application is that the acquired product-related information includes a vector indicating a browsing status of whether or not a same user has browsed each of a plurality of products on a website of a transaction service enabling purchase of a product and the product corresponding to the product-related information during the same period, and a vector of an image of the product corresponding to the product-related information, and the calculating means calculates the similarity based on the vector indicating the browsing status and the vector of the image. This is a recommended product determination device.

[0027] According to this aspect, the similarity is calculated based on the vector indicating the browsing status during the same period of each of the purchased product and related products with other products and the vector of the image. Therefore, the similarity between the purchased product and the related products can be obtained more appropriately.

[0028] Another aspect of the present application is that the calculating means integrates the vectors included in the acquired product-related information for each of the purchased product and the related products to generate a vector for similarity calculation, and calculates the similarity of the vector for similarity calculation. This is a recommended product determination device.

[0029] According to this aspect, by integrating the vectors, the similarity can be obtained by a single calculation.

[0030] Another aspect of the present application is that the products having the predetermined relationship include products purchased by the same user during the same period as the products indicated by the acquired purchase history information, and the specifying means is second purchase history information indicating purchase histories of products by a plurality of users, and specifies the products having the predetermined relationship based on the second purchase history information indicating the purchased product, the time when the product was purchased, and the user who purchased the product. This is a recommended product determination device.

[0031] According to this perspective, related products include items purchased by the same user at the same time as the items purchased by the target user. The target user may also desire items purchased at the same time as the items they purchased. These items may or may not be similar. Therefore, related products can include items that are relatively dissimilar to the purchased items, as these are items the target user may desire.

[0032] Another aspect of the present invention is a recommended product determination device characterized in that the conditions for the recommended product include a condition regarding the age of the purchase date, and the determination means increases the degree to which the condition regarding the age of the purchase date is easily satisfied as the similarity decreases.

[0033] According to this aspect, the lower the degree of similarity, the higher the probability that the related product will become a recommended product.

[0034] Another aspect of the present invention is a recommended product determination device characterized in that the condition for the recommended product is that the purchase date is older than the standard age, and the determination means corrects the purchase date age so that the lower the similarity, the older the purchase date becomes, or determines the standard age so that the standard age becomes newer.

[0035] According to this perspective, the lower the degree of similarity, the easier it is to satisfy the conditions for similarity, and the older the purchase date, the higher the probability of satisfying those conditions.

[0036] Another aspect of the present invention relates to a computer-based method for determining recommended products, comprising: a purchase history information acquisition step, which acquires purchase history information indicating the purchase history of products by a user of a terminal device, the purchased products and the time when the products were purchased; a identification step, which identifies related products having a predetermined relationship with the purchased products indicated by the acquired purchase history information; a product relationship information acquisition step, which acquires product relationship information relating to products, the product relationship information of the purchased products and the product relationship information of the identified related products; and the product relationship information of the purchased products. The method for determining a recommended product includes: a calculation step of calculating the similarity between relationship information and the product relationship information of the identified related product; a determination step of determining a recommended product from among the identified related products that satisfies the conditions for a product to be recommended to the user, based on the age of the purchase date of the purchased product and the calculated similarity; and a recommended product information transmission step of transmitting recommended product information indicating the determined product to the terminal device, wherein the conditions for the recommended product include a condition for either the similarity or the age of the purchase date, and the determination step is characterized in that the degree of satisfaction of the condition for one of the two is changed according to the other.

[0037] Another aspect of the present invention is a computer comprising: purchase history information acquisition means for acquiring purchase history information indicating the purchase history of goods by a user of a terminal device, which includes the purchased goods and the time when the goods were purchased; identification means for identifying related goods that have a predetermined relationship with the purchased goods indicated by the acquired purchase history information; product relationship information acquisition means for acquiring product relationship information relating to goods, which includes the product relationship information of the purchased goods and the product relationship information of the identified related goods; and the product relationship information of the purchased goods and the identified related goods. The recommended product determination program functions as a calculation means for calculating the similarity of a product to the aforementioned product-related information, a determination means for determining a recommended product from among the identified related products that satisfies the conditions for a product to be recommended to the user, based on the age of the purchase date of the purchased product and the calculated similarity, and a recommended product information transmission means for transmitting recommended product information indicating the determined product to the terminal device, wherein the conditions for the recommended product include a condition for either the similarity or the age of the purchase date, and the determination means changes the degree of ease of satisfying the condition for one of them according to the other. [Effects of the Invention]

[0038] According to the present invention, it is possible to recommend products that suit the user's preferences, taking into account when the user purchases the product. [Brief explanation of the drawing]

[0039] [Figure 1] This figure shows an example of the general configuration of a communication system S according to one embodiment. [Figure 2] This is a block diagram showing an example of the outline configuration of an online shopping server 1 according to one embodiment. [Figure 3] This figure shows an example of the contents stored in the database of online shopping server 1. [Figure 4] This figure shows an example of a functional block of the system control unit 11 of an online shopping server 1 according to one embodiment. [Figure 5] (a) is a diagram showing an example of vector merging. (b) is a diagram showing an example of how vector similarity is calculated. [Figure 6] (a) is a graph showing an example of the relationship between the purchase date and the correction factor. (b) is a diagram showing an example of a calculation for determining the recommended product. [Figure 7] (a) is a diagram showing an example of how recommended products are determined. (b) is a diagram showing another example of how recommended products are determined. [Figure 8] This figure shows another example of how to determine recommended products. [Figure 9] This flowchart shows an example of a top page transmission process executed by the system control unit 11 of the online shopping server 1 according to one embodiment. [Figure 10] This flowchart shows an example of the recommended product determination process executed by the system control unit 11 of the online shopping server 1 according to one embodiment. [Modes for carrying out the invention]

[0040] A first embodiment of the present invention will be described in detail below with reference to the drawings.

[0041] [1. Communication System Configuration] First, the configuration and functional overview of the communication system S according to this embodiment will be explained with reference to Figure 1. Figure 1 is a diagram showing an example of the overview configuration of the communication system S according to this embodiment.

[0042] As shown in Figure 1, the communication system S consists of an online shopping server 1 and multiple user terminals 2. These devices are connected to a network NW. The network NW is constructed using, for example, the internet, dedicated communication lines (e.g., CATV (Community Antenna Television) lines), mobile communication networks (including base stations, etc.), and gateways.

[0043] Online shopping server 1 may be a server device that performs processing related to transaction services. Transaction services may be, for example, services that enable users to purchase goods in e-commerce. In transaction services, it may be possible to purchase goods from only one seller, or it may be possible to purchase goods from multiple sellers. Sellers are those who sell goods using the transaction services. Examples of sellers include companies, stores, and individuals. Transaction services may be provided, for example, on a designated website. Examples of such websites include online shopping sites and online shopping malls. Hereafter, as an example, a case in which multiple stores sell goods in an online shopping mall will be described.

[0044] Online shopping server 1 may send the content of the trading service website to user terminal 2 in response to a request from user terminal 2. Online shopping server 1 may also perform processes such as searching for products, sending search results, sending information on products selected by the user, and accepting product orders. Furthermore, online shopping server 1 may be an example of a recommended product determination device. A recommended product determination device may be a device that determines which products to recommend to the user from among the products available for purchase through the trading service. These products are called recommended products. Recommended products may be determined individually for each user. Online shopping server 1 may determine recommended products based, for example, on purchase history information. Purchase history information may be information indicating the history of product purchases on the trading service. Online shopping server 1 may determine recommended products from among products that have some relationship with products purchased by the user, for example. Online shopping server 1 may send information indicating the determined recommended products to user terminal 2.

[0045] Each user terminal 2 may be a terminal device used by a user who is eligible to use the trading service. Examples of user terminals 2 include personal computers, smartphones, other mobile phones, tablet computers and other portable information terminals, PDAs (Personal Digital Assistants), set-top boxes, etc. A web browser may be installed on each user terminal 2. Users may be able to use the trading service by having user terminal 2 receive and display the trading service content from the online shopping server 1 via the web browser. Certain types of user terminals may be capable of having an application for the trading service installed. Users may be able to use the trading service through this application.

[0046] [2. Online Shopping Server Configuration] Next, the configuration of the online shopping server 1 will be described with reference to Figures 2 and 3. Figure 2 is a block diagram showing an example of the schematic configuration of the online shopping server 1 according to this embodiment. As shown in Figure 2, the online shopping server 1 includes a system control unit 11, a system bus 12, an input / output interface 13, a storage unit 14, and a communication unit 15. The system control unit 11 and the input / output interface 13 are connected via the system bus 12.

[0047] The system control unit 11 is composed of a CPU (Central Processing Unit) 11a, a ROM (Read Only Memory) 11b, a RAM (Random Access Memory) 11c, and the like.

[0048] The input / output interface 13 performs interface processing between the storage unit 14 and the communication unit 15 and the system control unit 11.

[0049] The storage unit 14 is composed of, for example, a hard disk drive. This storage unit 14 may store databases such as product DB 14a, browsing history DB 14b, purchase history DB 14c, product vector DB 14d, and concurrently purchased product DB 14e. "DB" is an abbreviation for database.

[0050] Figure 3 shows an example of the contents stored in the database of the online shopping server 1. The product DB 14a may store product information as information about products sold in the trading service, for each combination of product and the store that sells that product. For example, as shown in Figure 3, the product DB 14a may store product information such as store ID, product ID, product code, category information, product name, catchphrase, product description, product image URL (Uniform Resource Locator), price, etc., in relation to each other. The store ID may be identification information for identifying the store that sells the product. The product ID and product code may each be information for identifying the product. The product ID may be identification information assigned by the store that sells the product to identify that product. The product code may be unique identification information predetermined for the product, regardless of the store that sells it. An example of a product code is the JAN (Japanese Article Number) code. In this embodiment, the combination of store ID and product ID corresponds to product identification information for identifying a product. The product identification information makes it possible to identify which product is sold at which store. However, the product ID alone may constitute product identification information. Category information may indicate the category to which the product belongs. Product categories may be defined hierarchically, for example. Examples of top-level categories include fashion, food, beverages, daily necessities, cosmetics, electrical appliances, computers, sports, interior design, games, books, etc. Examples of categories belonging to fashion include women's fashion, men's fashion, kids' and baby fashion, underwear, etc. Product category information may include, for example, the category IDs of each category from the highest-level category to which the product belongs down to the lowest-level category. The category ID may also be identification information for identifying a category. The product name may be the name given to the product by the store that sells it. The product name stored in product DB14a may differ from the official name of the product.Generally, product names often include the official name of the product, but they may also include other wording. For example, a product name may include supplementary information, descriptions, or promotional phrases about the product. A catchphrase may be a promotional phrase specified by the selling store for the product. A product description may be a description of the product specified by the selling store. The product name, catchphrase, and product description may each be text data. A product image URL may be identification information for identifying a product image. A product image may be an image of the product. Image data for each product image may be stored separately in storage unit 14 in association with the product image URL. The product name, catchphrase, product description, product image, and price may be information presented to the user on the trading service website.

[0051] Each piece of information included in the product information can be entered by store employees using a terminal device (not shown) located in the store. The online shopping server 1 generates product information based on the information entered by the employees and stores this product information in the product database 14a. The online shopping server 1 also receives image data of product images from the terminal device and stores the image data in the storage unit 14, associating it with the product image URL.

[0052] When product information is stored in product DB14a, or when product information stored in product DB14a is updated, online shopping server 1 may generate an HTML document for the product page for the registered product based on this product information. The product page may be a web page corresponding to the screen of a specific product. The product page may be a web page that contains information about that product. For example, the product page may include the product name, catchphrase, product description, product image, price, etc. The product page may also be a screen that accepts operations for purchasing the product that the product page corresponds to. For example, the product page may include elements that can be operated on to purchase the product. Examples of operable elements include buttons, links, icons, lists, etc. Examples of operations for purchasing a product include operations to enter the purchase procedure and operations to add the product to the shopping cart. The purchase procedure may include processes for the user to decide on the order details or processes for the user to confirm the order details. Examples of order details include the number of products to purchase, payment method, product delivery method, product delivery address, etc. The purchase procedure may further include processes to confirm the order details. The purchase procedure may also include a payment process in which the user pays the price for the goods purchased by the online shopping server 1. The purchase procedure may also include a process in which the online shopping server 1 notifies the store where the purchase is to be made of the order details. The shopping cart may be a virtual container or list in which the user places items that they plan to purchase later. The user can purchase the items in the shopping cart. The online shopping server 1 may store the generated product page HTML document in the storage unit 14, associating it with the store ID and product ID that the product page corresponds to. For example, the URL of the product page HTML document may include the store ID and product ID. The product page HTML document may be associated with this URL.

[0053] The browsing history DB14b may be a database that shows the browsing history of web pages in the trading service. In particular, the browsing history DB14b may show the browsing history of the product pages mentioned above. The browsing history DB14b may store a browsing log that shows the record of each time a product page is viewed. For example, the browsing history DB14b may store the browsing log as a browsing log, with the browsing date and time, user ID, store ID, and product ID associated with each other. The browsing date and time may indicate the date and time when the product page was viewed. The user ID may be identification information that identifies the user who viewed that product page. The store ID may indicate the store that sells the product on which the product page was viewed. The product ID may indicate the product on which the product page was viewed.

[0054] The purchase history DB14c may be a database that shows the history of product purchases in the trading service. The purchase history DB14c may store a purchase log each time a product is purchased. For example, the purchase history DB14c may store the order number, order date and time, user ID, and information on one or more purchased products as a purchase log, associated with each other. The order number may be a number that identifies the order for a product. The order date and time may indicate the date and time when the order for the product was placed. The order date and time may correspond to the date and time when the product was purchased. The user ID may indicate the user who purchased the product. The purchased product information may indicate the product that was purchased. If multiple products are purchased at once, the purchase log may include information on multiple purchased products. The purchased product information may include, for example, a store ID, product ID, price, and quantity purchased. The store ID may indicate the store that sold the purchased product. The product ID may indicate the purchased product. The price may indicate the unit price of the purchased product. The quantity purchased may indicate how many of that product were purchased.

[0055] Product Vector DB14d may store vectors of product-related information for each product and the store that sells it, for each combination of product and the store that sells it. Product-related information may be information related to a product. Product-related information may be any of the information contained in the product information stored in Product DB14a, or it may be other information related to a product. The vector of product-related information may be information that shows a distributed representation of product-related information by embedding, information that shows features or feature quantities of product-related information, or other information generated by vectorizing product-related information. The vector of product-related information may also be considered as a vector of a product. Generally, the vector of product-related information may be information used to determine the similarity between products. Product Vector DB14d may store, for example, a store ID, a product ID, and one or more types of vectors. The store ID may indicate the store that sells the product to which the vector corresponds. The product ID may indicate the product to which the vector corresponds. Examples of vectors stored in Product Vector DB14d include text vectors, image vectors, and concurrent viewing status vectors. The product vector DB14d may store one or two predetermined vectors, or all of them, from among text vectors, image vectors, and simultaneous viewing status vectors.

[0056] A text vector may be a vector representing text features or distributed representations of a target product identified by a combination of store ID and product ID. This text may be text viewable by users on the trading service website. Examples of this text include the product name, catchphrase, and product description included in the product information. The text vector may be a vector of one predetermined text from among the product name, catchphrase, and product description. Alternatively, the text vector may be a vector of text generated by combining two predetermined texts or all of these texts. Examples of models used to generate text vectors include Word2vec, Doc2Vec, and BERT (Bidirectional Encoder Representations from Transformers). An example of a method for generating text vectors is the use of an embedded API (Application Programming Interface) provided by OpenAI®. For example, when the online shopping server 1 stores product information in the product DB 14a, or updates product information stored in the product DB 14a, it may generate a text vector using the text included in that product information. The online shopping server 1 may store the generated text vectors in the product vector DB 14d, associating them with the store ID and product ID.

[0057] The image vector may be a vector representing the features or distributed representation of the product image of a target product identified by a combination of store ID and product ID. An example of a model used to generate the image vector is ViT (Vision transformer). Another example of a method for generating the image vector is to use the embedded API provided by Google® Gemini. For example, when the online shopping server 1 stores image data of a new product in the storage unit 14, it may generate an image vector using that image data. The online shopping server 1 may store the generated image vector in the product vector DB 14d, associated with the store ID and product ID.

[0058] The simultaneous browsing status vector may also be a vector of simultaneous browsing status information. Simultaneous browsing status information may be information indicating whether the same user viewed each of multiple products available for purchase or viewing on the trading service website, and a target product identified by a combination of store ID and product ID, at the same time. Specifically, simultaneous browsing status information may be information indicating whether the same user viewed each of multiple products' product pages and the target product's product page at the same time. The simultaneous browsing status vector may also be a vector indicating this browsing status. Simultaneous browsing may mean that a user views the product page of one product and then views the product page of another product within a predetermined time. Examples of predetermined time include 10 minutes, 30 minutes, 1 hour, 3 hours, etc. Simultaneous browsing status information may be generated using collaborative filtering based on, for example, the browsing history DB 14b. For example, the online shopping server 1 obtains the browsing date and time, user ID, store ID, and product ID contained in a browsing log stored in the browsing history DB 14b. The product indicated by the obtained combination of store ID and product ID is designated as the first product. Online shopping server 1 searches the browsing history DB 14b for browsing logs that include browsing dates and times within a predetermined time before and after the acquired browsing date and time, and that contain the same user ID as the acquired user ID. The product indicated by the combination of store ID and product ID included in the retrieved browsing log is designated as the second product. Online shopping server 1 may determine that the first product and the second product were viewed by the same user at the same time. For each browsing log stored in the browsing history DB 14b, online shopping server 1 may identify products viewed by the same user at the same time as the product corresponding to that browsing log. Online shopping server 1 may or may not generate a product-product matrix. Each row of the product-product matrix is ​​associated with a different product that is available for purchase or viewing in the transaction service. Each column of the product-product matrix is ​​also associated with a different product that is available for purchase or viewing in the transaction service.Each element of the product-product matrix may indicate whether the product associated with the row containing that element and the product associated with the column containing that element were viewed simultaneously by the same user. For example, if those products were viewed simultaneously, the element may be set to 1, and if they were not viewed simultaneously, the element may be set to 0. Alternatively, each element of the product-product matrix may indicate the number of users who viewed the product associated with the row containing that element and the product associated with the column containing that element simultaneously. The simultaneous viewing status information may correspond to the information of the row or column to which the target product is associated. The simultaneous viewing status information itself may be used as the simultaneous viewing status vector. However, since the number of products available for purchase in a trading service is generally substantial, the dimensionality of the simultaneous viewing status vector can become enormous. Therefore, the online shopping server 1 may reduce the dimensionality of the simultaneous viewing status vector or generate a simultaneous viewing status vector with a dimensionality considerably less than the number of products. Examples of methods for generating such vectors include Matrix Factorization and Factorization Machine. The online shopping server 1 may, for example, periodically update the concurrent browsing status vector for each product stored in the product vector DB 14d based on the latest browsing history DB 14b. Examples of intervals for updating the concurrent browsing status vectors include one day, one week, one month, etc.

[0059] The concurrently purchased items DB14e may store information indicating products purchased by the same user at the same time as each product sold in the trading service. For example, the concurrently purchased items DB14e may store at least a store ID and a product ID associated with each other. The store ID may indicate the store that sold the product in question. The product ID may indicate the product in question. If there are products purchased at the same time as the product in question, the concurrently purchased items DB14e may store one or more pieces of information on concurrently purchased products associated with the store ID and product ID. The concurrently purchased product information may indicate products purchased at the same time as the product in question. For example, the concurrently purchased product information may include a store ID and a product ID. The store ID may indicate the store that sold the products purchased at the same time. The product ID may indicate the products purchased at the same time.

[0060] The online shopping server 1 may, for example, periodically identify products purchased by the same user at the same time as a product, based on the latest purchase history DB 14c. The online shopping server 1 may, for example, use collaborative filtering. For example, the online shopping server 1 may determine that only products purchased at the same time as a product were purchased by the same user at the same time. If the purchase log stored in the purchase history DB 14c contains information on multiple purchased products, the online shopping server 1 can determine that the products indicated by that purchase information were purchased simultaneously by the same user. Alternatively, the online shopping server 1 may determine that products purchased within a predetermined time after a certain product was purchased are products purchased at the same time as that product. Examples of predetermined time include 1 hour, 6 hours, 1 day, etc. In this case, the online shopping server 1 retrieves the order date and time, user ID, and purchased product information contained in a purchase log stored in the purchase history DB 14c. The product indicated by the retrieved purchased product information is designated as the first product. Online shopping server 1 searches the purchase history DB 14c for browsing logs that include order dates and times within a predetermined time before and after the acquired order date and time, and that include the same user ID as the acquired user ID. The product indicated by the purchased product information included in the retrieved purchase log is designated as the second product. Online shopping server 1 may determine that the first product and the second product were purchased by the same user at the same time. Online shopping server 1 may generate a product-product matrix in a manner similar to the method described above. Each element of the product-product matrix may indicate the number of users or purchase counts that purchased the product associated with the row containing that element and the product associated with the column containing that element at the same time. Online shopping server 1 may, for example, divide the number of purchasers or purchase counts of each element included in the row by the number of users who purchased the product associated with that row or the number of times that product was purchased. In this way, online shopping server 1 may calculate the proportion of each product that was purchased at the same time as the product associated with that row.Online shopping server 1 may ultimately determine that products with a ratio greater than or equal to a predetermined value were purchased at the same time as the products associated with that row. Online shopping server 1 may associate the store ID and product ID of the products associated with that row with the information of products purchased at the same time, indicating the products that were determined to have been purchased at the same time, and store this information in the concurrently purchased products DB 14e.

[0061] Furthermore, the storage unit 14 may store various programs such as an operating system, a DBMS (Database Management System), and a trading service program. The trading service program is a program that causes the system control unit 11 to execute processing related to trading services. The trading service program may be acquired, for example, from another device via a network NW, or it may be recorded on a recording medium such as magnetic tape, optical disk, or memory card and read via a drive device.

[0062] The communication unit 15 is composed of, for example, a network interface card. The communication unit 15 connects to a different device from the online shopping server 1 via the network NW and controls the communication status with the connected device.

[0063] [3. Overview of System Control Unit Functions] Next, the functional overview of the system control unit 11 of the online shopping server 1 will be described with reference to Figures 4 to 7. Figure 4 is a diagram showing an example of the functional blocks of the system control unit 11 of the online shopping server 1 according to this embodiment. The system control unit 11 may function as a purchase history acquisition unit 1101, a related product identification unit 1102, a product related information acquisition unit 1103, a similarity calculation unit 1104, a recommended product determination unit 1105, and a recommended product information transmission unit 1106, etc., as shown in Figure 4, by the CPU 11a reading and executing various program codes included in the transaction service program.

[0064] The purchase history acquisition unit 1101 may acquire purchase history information showing the purchase history of a product by the target user. The target user may be a user to whom the product is recommended. The target user is called the recommended user. For example, in response to an operation by the user, the user terminal 2 may send a request to the online shopping server 1 for an HTML document of a web page containing information on recommended products on the trading service's website. The user of the user terminal 2 that sent this request may be the recommended user. The web page containing information on recommended products may be, for example, the top page. The top page may be the web page that serves as the entrance to the trading service's website. On the top page, the user may be able to perform operations to search for products. Alternatively, the online shopping server 1 may determine the recommended user based on some criteria. The purchase history acquisition unit 1101 may acquire purchase history information in response to a request from the user terminal 2. The purchase history information may show the purchased product and the time when the product was purchased. The purchase history information may also show the user who purchased the product. The purchase history acquisition unit 1101 may acquire purchase history information from, for example, the purchase history DB 14c. For example, the purchase history acquisition unit 1101 may search for one or more purchase logs containing the user ID of the recommended user as purchase history information. The combination of store ID and product ID included in the purchase log indicates the purchased product. The order date and time included in the purchase log indicates the time of purchase of the product. The purchase history acquisition unit 1101 may acquire all purchase logs for the recommended user. Alternatively, the purchase history acquisition unit 1101 may acquire only purchase logs that indicate the purchase time within a period starting from a predetermined length prior to today and ending today.

[0065] The related product identification unit 1102 may identify related products that have a predetermined relationship with the product indicated by the purchase history information acquired by the purchase history acquisition unit 1101 as a product purchased by the recommended user. Related products may include related products related to the product purchased by the recommended user. Products that have a predetermined relationship with the purchased product may be of a different type than the relationship indicated by the similarity of product relationship information between the purchased product and the product, as described later.

[0066] Related products may include products purchased by the same user at the same time as the product purchased by the recommended user. In this case, the related product identification unit 1102 may search the concurrently purchased products DB 14e for concurrently purchased products associated with the store ID and product ID combination of the product purchased by the recommended user. The related product identification unit 1102 may identify the product indicated by the retrieved concurrently purchased products as a product purchased by the same user at the same time as the product purchased by the recommended user. Alternatively, the related product identification unit 1102 may search the purchase history DB 14c for purchase logs containing purchase product information, including the store ID and product ID combination of the product purchased by the recommended user, regardless of who the user who purchased the product was. The related product identification unit 1102 may identify the product indicated by other purchase product information included in the retrieved purchase logs as a product purchased by the same user at the same time as the product purchased by the recommended user. The related product identification unit 1102 may also search the purchase history DB 14c for purchase logs that include the order date and time within the predetermined period from the order date and time included in the searched purchase log, and that include the same user ID as the user ID included in the searched purchase log. The related product identification unit 1102 may identify the products indicated by the purchased product information included in the searched purchase log as products purchased by the same user at the same time as the products purchased by the recommended user. For each product purchased by the same user at the same time as the product purchased by the recommended user, the related product identification unit 1102 may divide the number of users who purchased at the same time or the number of times the product was purchased at the same time by the number of users who purchased the product purchased by the recommended user or the number of times that product was purchased. By doing so, the related product identification unit 1102 may calculate the proportion of each product that was purchased at the same time as the product purchased by the recommended user. The related product identification unit 1102 may include only products for which this proportion is equal to or greater than a predetermined value as related products.

[0067] Related products may be products that match specific types of information with products purchased by the recommended user. For example, related products may include products belonging to the same category as products purchased by the recommended user. In this case, the related product identification unit 1102 may obtain category information of products purchased by the recommended user from the product DB 14a. The related product identification unit 1102 may search the product DB 14a for product information that includes the same category information as the obtained category information. The related product identification unit 1102 may identify the product corresponding to the searched product information as a product belonging to the same category as products purchased by the recommended user. Alternatively, the related product identification unit 1102 may obtain the category ID of a predetermined hierarchical category included in the obtained category information. The related product identification unit 1102 may search the product DB 14a for product information that includes the same category ID as the obtained category ID in its category information. The related product identification unit 1102 may identify the product corresponding to the searched product information as a product belonging to the same category as products purchased by the recommended user. In this case, the specific type of information is the category ID or category information.

[0068] Related products may include products of the same brand or series as the product purchased by the recommended user. For example, the storage unit 14 may store a catalog DB (not shown). The catalog DB may be a database showing product catalogs. For example, the catalog DB may store information about each product in association with its product code. For at least some products, the catalog DB may store at least one of the following: identification information for identifying the brand of the product and identification information for identifying the series of the product. The related product identification unit 1102 may obtain the product code of the product purchased by the recommended user from the product DB 14a. The related product identification unit 1102 may obtain the brand or series identification information of the product based on the obtained product code and the catalog DB. The related product identification unit 1102 may search the catalog DB for a product code associated with the same identification information as the retrieved product code. The related product identification unit 1102 may search the product DB 14a for product information including the retrieved product code. The related product identification unit 1102 may identify the product corresponding to the searched product information as a product of the same brand or series as the product purchased by the recommended user. In this case, the type of information identified is brand identification information or series identification information.

[0069] Multiple predetermined relationships may be defined. For example, related products may include products purchased by the same user at the same time, products in the same category, and products of the same brand or series. The related product identification unit 1102 may identify products as related products if they have at least one of the predetermined relationships with the purchased product.

[0070] The product-related information acquisition unit 1103 may acquire product-related information about products. In particular, the product-related information acquisition unit 1103 may acquire product-related information for products indicated by purchase history information acquired by the purchase history acquisition unit 1101 as products purchased by recommended users, and product-related information for related products identified by the related product identification unit 1102. The product-related information may include any of the information contained in the product information stored in the product DB 14a. For example, the product-related information may include text viewable on the trading service website. Examples of this text include product name, catchphrase, and product description. The product-related information may also include image data of product images stored in the storage unit 14. The product-related information may also include information about user behavior towards products on the trading service website. For example, the behavior information may include information about product browsing. For example, information about product browsing may include simultaneous browsing status information. The product-related information may include product vectors. Examples of product vectors include text vectors, image vectors, and simultaneous browsing status vectors. The product-related information may include multiple types of vectors. For example, the product-related information may include text vectors and image vectors. Alternatively, the product-related information may include text vectors and concurrent viewing status vectors. Alternatively, the product-related information may include image vectors and concurrent viewing status vectors. Alternatively, the product-related information may include text vectors, image vectors, and concurrent viewing status vectors. The product-related information acquisition unit 1103 may acquire product vectors from the product vector DB 14d. Alternatively, the product-related information acquisition unit 1103 may generate product vectors using the method described above.

[0071] The similarity calculation unit 1104 may calculate the similarity between the product relationship information of a product purchased by the recommended user and the product relationship information of a related product, which is obtained by the product relationship information acquisition unit 1103. This similarity may represent the similarity between the purchased product and the related product. The method of calculating the similarity is not particularly limited. If a product vector is obtained as product relationship information, the similarity calculation unit 1104 may calculate cosine similarity. If multiple types of vectors are obtained as product relationship information by the product relationship information acquisition unit 1103, the similarity calculation unit 1104 may calculate the similarity based on the multiple types of vectors. For example, the similarity calculation unit 1104 (or the product relationship information acquisition unit 1103) may integrate or combine multiple types of vectors for each of the products purchased by the recommended user and the related products to generate a vector for similarity calculation. For example, the product relationship information acquisition unit 1103 may merge or mix multiple types of vectors. Vector merging may, for example, generate a new vector by alternating the elements of each vector according to the order in which their elements are arranged. Vector merging may also be a type of vector integration or combination. For example, let i be the index of an element in the first vector and j be the index of an element in the second vector. i and j are both non-negative integers. The i-th element of the first vector becomes the i × 2 element in the generated similarity calculation vector. The j-th element of the second vector becomes the j × 2 + 1 element in the generated similarity calculation vector. Figure 5(a) shows an example of vector merging. As shown in Figure 5(a), suppose we merge a text vector V1 and an image vector V2. Text vector V1 contains elements V1-1, V1-2, V1-3, V1-4, and V1-5 in that order. The image vector V2 contains elements V2-1, V2-2, V2-3, V2-4, and V2-5 in that order. The vector V3 for similarity calculation generated by the merge contains elements V1-1, V2-1, V1-2, V2-2, V1-3, V2-3, V1-4, V2-4, V1-5, and V2-5 in that order.Furthermore, if there is a regularity in the correspondence between the positions of the elements of each vector before merging and the positions of the elements after merging, the merging method is not limited to the method described above. Also, the similarity calculation unit 1104 may simply combine the first vector and the second vector in series. The similarity calculation unit 1104 may also calculate the similarity of the vectors for similarity calculation generated by the integration. Figure 5(b) shows an example of how the similarity of vectors is calculated. As shown in Figure 5(b), the similarity between the vector V4 for similarity calculation of products purchased by the recommended user and the vector V5 for similarity calculation of related products is calculated.

[0072] When integrating three or more types of vectors, the similarity calculation unit 1104 may, for example, first merge two of those vectors. Then, the similarity calculation unit 1104 may merge the vector generated by the merge with the remaining vector to generate a vector for similarity calculation.

[0073] The vectors for similarity calculation may be generated in advance and stored in the product vector DB 14d. For example, the product relationship information acquisition unit 1103 may generate the vectors for similarity calculation as product relationship information from the product vector DB 14d. Furthermore, if the product relationship information includes text or images, the similarity calculation unit 1104 may generate vectors of that text or image.

[0074] The recommended product determination unit 1105 may determine a recommended product that satisfies the conditions for a product to be recommended to a user, based on the age of the purchase of the product indicated by the purchase history information obtained by the purchase history acquisition unit 1101 as a product purchased by the recommended user, and the similarity calculated by the similarity calculation unit 1104. The age of the purchase may be indicated, for example, by the length of time that has elapsed from the time the product was purchased until today. The longer this elapsed period, the older the purchase date. Examples of units for the length of the elapsed period include days, weeks, months, and years. The purchase date may also be indicated by the order date and time in the purchase log included in the purchase history information.

[0075] The conditions under which a related product becomes a recommended product are called recommendation conditions. Recommendation conditions may include conditions for either the similarity of product-related information or the age of the purchase date of the product purchased by the recommended user. The recommended product determination unit 1105 may change the degree of satisfaction of one of the conditions, similarity or age of purchase date, depending on the other. If the recommendation conditions include conditions for similarity, the recommended product determination unit 1105 may change the degree of satisfaction of that condition depending on the age of the purchase date. For example, the recommended product determination unit 1105 may change or determine the degree of satisfaction of the similarity condition calculated between the product and the related product, depending on the age of the purchase date of the product purchased by the recommended user. If the recommendation conditions include conditions for age of purchase date, the recommended product determination unit 1105 may change the degree of satisfaction of that condition depending on the similarity. For example, the recommended product determination unit 1105 may, depending on the similarity, change or determine the degree to which the condition regarding the age of purchase of the purchased product is easily satisfied for the related product for which the similarity was calculated. The degree to which the condition is easily satisfied may indicate the degree to which the condition is easily satisfied. The higher the degree to which the condition is easily satisfied, the easier it is for the similarity or the age of purchase to satisfy that condition. The degree to which the condition is easily satisfied may also correspond to the degree of strictness of that condition. The higher the degree of strictness, the harder it is for the similarity or the age of purchase to satisfy that condition. In this way, regardless of whether there is a condition for similarity or purchase date, the recommended product can be determined according to the user's preferences by considering the timing of the user's purchase of the product.

[0076] The following describes a case where the recommendation criteria include conditions for similarity. The recommendation product determination unit 1105 may increase the ease of satisfying the conditions as the purchase date is older. That is, the recommendation product determination unit 1105 may make it easier for similarity to satisfy the conditions as the purchase date is older. This may correspond to the conditions being more relaxed as the purchase date is older, or to the conditions being strengthened as the purchase date is newer.

[0077] The recommendation condition may be that the comparison between the similarity calculated by the similarity calculation unit 1104 and a reference value for that similarity results in a predetermined outcome. The reference value may be a threshold. In this case, the recommendation product determination unit 1105 may correct the similarity based on the age of the purchase. For example, the recommendation product determination unit 1105 may correct the similarity so that the older the purchase date, the more likely the comparison between the corrected similarity and the reference value will result in a predetermined outcome. The recommendation product determination unit 1105 may then determine the recommended product by determining whether the comparison between the similarity and the reference value results in a predetermined outcome, based on the corrected similarity and a predetermined reference value. Alternatively, the recommendation product determination unit 1105 may determine the reference value based on the age of the purchase. In this case, the reference value may correspond to the degree of ease of satisfying the similarity condition. For example, the recommendation product determination unit 1105 may determine the reference value so that the older the purchase date, the more likely the comparison between the similarity and the reference value will result in a predetermined outcome. The recommended product determination unit 1105 may then determine the recommended product by determining whether the result of comparing the similarity and the determined standard value is a predetermined result, based on the similarity and the determined standard value.

[0078] The recommended product determination unit 1105 may determine a correction coefficient for similarity based on the age of the purchase date. The correction coefficient may be a coefficient for correcting similarity. In this case, the correction coefficient may correspond to the degree to which the similarity conditions are easily satisfied. For example, the recommended product determination unit 1105 may determine the correction coefficient such that the older the purchase date, the more likely the comparison between similarity and the baseline value is to yield a predetermined result. The correction coefficient may be determined, for example, between 0 and 1. The recommended product determination unit 1105 may calculate the correction coefficient from the age of the purchase date using a predetermined calculation formula. Alternatively, the recommended product determination unit 1105 may determine the correction coefficient based on a table showing the correspondence between the age of the purchase date and the correction coefficient. The recommended product determination unit 1105 may calculate the corrected similarity based on the determined correction coefficient and similarity. The recommended product determination unit 1105 may calculate the corrected similarity by multiplying the similarity by the correction coefficient.

[0079] The recommendation condition may be, for example, that the similarity calculated by the similarity calculation unit 1104 is below a threshold value. In this case, the recommendation product determination unit 1105 may adjust the similarity so that the adjusted similarity becomes lower the older the purchase date. The lower the similarity, the easier it is to satisfy the condition for similarity. For example, the recommendation product determination unit 1105 may make the adjustment coefficient smaller the older the purchase date. The smaller the adjustment coefficient, the easier it is to satisfy the condition for similarity. Alternatively, the recommendation product determination unit 1105 may make the threshold value larger the older the purchase date. The higher the threshold value, the easier it is to satisfy the condition for similarity.

[0080] Users purchase products because they desire them. Once a product is purchased, the user's desire for that product is satisfied. Therefore, users who have just purchased a product are unlikely to want the same or similar product. On the other hand, assuming that the product purchased by a user suits that user's taste, similar products may also suit that user's taste. Therefore, as time passes after a user purchases a product, they tend to want to purchase the same or similar product. For example, a user may get tired of the product they purchased and want a new one. Or, a user may want to replace a product because it has become old or is nearing the end of its lifespan. According to the operation of the recommended product determination unit 1105, related products with a relatively high degree of similarity to the purchased product are less likely to be recommended, but as time passes since the purchase, they become more likely to be recommended. Therefore, it is possible to recommend products that meet the user's needs. Furthermore, related products with a relatively low degree of similarity to the purchased product are more likely to become recommended products, even if they were purchased relatively recently. Even with low similarity, related products have a predetermined relationship with the purchased product. Users tend to want products that are related to the products they have purchased. Moreover, because these related products have a low degree of similarity to the product that satisfied the user's purchase desire, the user is likely to want that product as well. If related products include products purchased by the same user at the same time as the product purchased by the recommended user, the recommended user is more likely to want to purchase related products for the product they recently purchased. In this way, by considering the timing of the user's purchase, it is possible to determine recommended products that match the user's preferences.

[0081] Figure 6(a) is a graph showing an example of the relationship between the age of purchase and the correction coefficient. In Figure 6(a), the age of purchase is indicated by the number of weeks that have passed since the product was purchased. As shown in Figure 6(a), the more weeks that have passed, i.e., the older the purchase date, the smaller the correction coefficient becomes. For example, the correction coefficient may decrease exponentially with increasing age of purchase. Alternatively, the correction coefficient may decrease in a different manner. Figure 6(b) is a diagram showing an example of a calculation for determining recommended products. As shown in Figure 6(b), the correction coefficient determined according to the age of purchase is multiplied by the similarity to calculate the corrected similarity. Related products whose similarity falls below the threshold value become recommended products.

[0082] Purchase history information may indicate that multiple products have been purchased by the recommended user. In this case, we will explain how to determine whether the recommendation conditions are met in relation to each of the purchased products. For example, the recommendation product determination unit 1105 may determine whether the recommendation conditions are met in relation to each of the purchased products for each related product. The recommendation product determination unit 1105 may then determine the related product that meets the recommendation conditions in relation to all the purchased products as the recommended product. In this case, the similarity calculation unit 1104 may calculate the similarity of the product relationship information for each related product in relation to each of the purchased products. For example, suppose the purchased products include at least a first product and a second product that is different from the first product. Each of the first and second products may be any of the purchased products. The similarity calculation unit 1104 may first calculate the similarity between the product relationship information of the related product and the product relationship information of the first product. This similarity is called the first similarity. The recommended product determination unit 1105 determines whether the related product meets the recommendation criteria based on the first similarity and the age of the first product's purchase. For example, the recommended product determination unit 1105 corrects the first similarity based on the age of the first product's purchase. The recommended product determination unit 1105 determines whether the result of comparing the corrected first similarity with a predetermined standard value is a predetermined result. If the comparison result is a predetermined result, the recommended product determination unit 1105 determines that the related product meets the recommendation criteria. In this case, the similarity calculation unit 1104 may calculate the similarity between the product relationship information of the related product and the product relationship information of the second product. This similarity is called the second similarity. The recommended product determination unit 1105 determines whether the related product meets the recommendation criteria based on the second similarity and the purchase date of the second product. Similarly thereafter, the similarity calculation unit 1104 and the recommended product determination unit 1105 may perform similarity calculations and determine whether the recommendation criteria are met for each of the purchased products. The recommended product determination unit 1105 may then determine that a related product that satisfies the recommendation criteria in relation to all the purchased products is a recommended product.On the other hand, if the system determines that a related product does not meet the recommendation criteria based on the first similarity and the age of the first product's purchase, the recommendation product determination unit 1105 may exclude the first product from the list of recommended products. In this case, the similarity calculation unit 1104 does not need to calculate the second similarity. The similarity calculation unit 1104 does not need to calculate the similarity of the related product with the second product and other products thereafter.

[0083] Figure 7(a) shows an example of how recommended products are determined. As shown in Figure 7(a), the recommended user has previously purchased products 110, 120, 130, and 140. Product 150 has also been identified as a related product to one of the purchased products. Assume that the similarity is adjusted based on the age of the product purchases. The baseline value is 0.5. For example, the similarity calculation unit 1104 calculates the adjusted similarity between related product 150 and product 110. The adjusted similarity is 0.2. Therefore, the recommendation condition is met. The similarity calculation unit 1104 then calculates the adjusted similarity between related product 150 and product 120. The adjusted similarity is 0.4. The similarity calculation unit 1104 then calculates the adjusted similarity between related product 150 and product 130. The adjusted similarity is 0.35. The similarity calculation unit 1104 then calculates the adjusted similarity between related product 150 and product 140. The corrected similarity score is 0.1. The corrected similarity score is below the threshold value for all purchased products. Therefore, the recommended product determination unit 1105 determines related product 150 as the recommended product.

[0084] Figure 7(b) shows another example of how recommended products are determined. As shown in Figure 7(b), the recommended user has previously purchased products 110, 120, 130, and 140. Product 160 has also been identified as a related product to one of the purchased products. The similarity calculation unit 1104 calculates the corrected similarity between related product 160 and product 110. The corrected similarity is 0.3. Therefore, the recommendation condition is met. The similarity calculation unit 1104 then calculates the corrected similarity between related product 150 and product 120. The corrected similarity is 0.8. Since the corrected similarity is higher than the baseline value, the recommendation condition is not met. Therefore, the recommended product determination unit 1105 excludes related product 150 from the recommended products. The similarity calculation unit 1104 does not calculate the similarity of related product 160 between products 130 and 140.

[0085] Next, we will explain an example of how to determine recommended products using specific products as examples. Figure 8 shows another example of how to determine recommended products. As shown in Figure 8, the recommended user has purchased products 210 and 220. Product 210 is a wristwatch. Product 210 was purchased relatively long ago. For example, product 210 was purchased 10 months ago. The correction factor for product 210 is 0.4. Product 220 is a smartphone. Product 220 was purchased relatively recently. For example, product 220 was purchased 3 days ago. The correction factor for product 220 is 1. Products 230, 240, and 250 were identified as related products. Product 230 is a related product of product 210. Product 230 is a wristwatch. The category to which product 210 belongs and the category to which product 230 belongs are the same. Products 240 and 250 are related products of product 220. Product 240 is a smartphone. The category to which product 220 belongs and the category to which product 240 belongs are the same. Product 250 is a smartphone strap. Product 250 is purchased by the same user at the same time as product 220. The similarity between product 230 and product 210 is 0.8. These products are of the same type, and functionally, they tend to be similar to each other. Therefore, the similarity tends to be relatively high. The correction factor is 0.4, so the corrected similarity is 0.32. The similarity between product 230 and product 220 is 0.2. These products are of different types, so the similarity tends to be relatively low. The correction factor is 1, so the corrected similarity is 0.2. If the baseline value is 0.5, product 230 is a recommended product. The similarity between product 240 and product 210 is 0.2. Therefore, the corrected similarity is 0.08. The similarity between product 240 and product 220 is 0.7. Therefore, the corrected similarity is 0.7. For this reason, product 240 is not a recommended product. The similarity between product 250 and product 210 is 0.3. Therefore, the corrected similarity is 0.12. The similarity between product 250 and product 220 is 0.4. Product 220 is a product related to smartphones. However, the shape of product 220 is significantly different from the shape of product 250. When comparing product images, the similarity tends to be low.On the other hand, when comparing products using text related to the product or products viewed at the same time, the similarity score may be higher or lower. The adjusted similarity score is 0.4. Therefore, product 240 is a recommended product.

[0086] Furthermore, the recommended product determination unit 1105 does not have to determine whether all of the purchased products meet the recommendation criteria. For example, it may only determine whether the related products meet the recommendation criteria for products that have a predetermined relationship with the other purchased products.

[0087] This section describes a case where the recommendation criteria include a criterion regarding the age of the purchase date. The recommendation product determination unit 1105 may increase the degree of ease of satisfying the criterion as the similarity decreases. That is, the recommendation product determination unit 1105 may make it easier to satisfy the criterion of the age of the purchase date as the similarity decreases. This may correspond to the criterion being more relaxed as the similarity decreases, or the criterion being more strengthened as the similarity increases.

[0088] The recommendation condition may be that the comparison between the age of the purchase date and the age of the standard results in a predetermined outcome. The age of the standard may be indicated by the length of time. In this case, the recommendation product determination unit 1105 may adjust the age of the purchase date based on the similarity calculated by the similarity calculation unit 1104. For example, the recommendation product determination unit 1105 may adjust the age of the purchase date so that the lower the similarity, the more likely the comparison between the age of the purchase date and the age of the standard will result in a predetermined outcome. The recommendation product determination unit 1105 may then determine the recommended product by determining whether the comparison between the age of the purchase date and the age of the standard results in a predetermined outcome, based on the age of the purchase date and the predetermined age of the standard. Alternatively, the recommendation product determination unit 1105 may determine the age of the standard based on similarity. In this case, the age of the standard may correspond to the degree to which the condition of the age of the purchase date is easily satisfied. For example, the recommended product determination unit 1105 may determine the age of the standard such that the lower the similarity, the more likely the comparison between the purchase date and the standard's age will result in a predetermined outcome. The recommended product determination unit 1105 may then determine the recommended product by determining whether the comparison between the purchase date and the standard's age will result in a predetermined outcome, based on the purchase date and the determined standard's age.

[0089] The recommended product determination unit 1105 may determine a correction coefficient for the age of the purchase date based on the similarity. The correction coefficient may be a coefficient for correcting the age of the purchase date. The correction coefficient may correspond to the degree to which the condition of the age of the purchase date is easily satisfied. For example, the recommended product determination unit 1105 may determine the correction coefficient such that the lower the similarity, the more likely the comparison between the age of the purchase date and the reference value is to result in a predetermined outcome. The correction coefficient may be determined, for example, between 0 and 1. The recommended product determination unit 1105 may calculate the corrected age of the purchase date based on the determined correction coefficient and the age of the purchase date. The recommended product determination unit 1105 may calculate the corrected age of the purchase date by multiplying the length of the period corresponding to the age of the purchase date by the correction coefficient.

[0090] The recommendation condition may be, for example, that the purchase date is older than the standard purchase date. In this case, the recommendation product determination unit 1105 may adjust the purchase date so that the older the adjusted purchase date, the lower the similarity. The older the purchase date, the easier it is to satisfy the condition regarding the purchase date. For example, the recommendation product determination unit 1105 may increase the adjustment coefficient as the purchase date gets older. The larger the adjustment coefficient, the easier it is to satisfy the condition regarding the purchase date. Alternatively, the recommendation product determination unit 1105 may make the standard purchase date newer as the similarity gets lower. The newer the standard purchase date, the easier it is to satisfy the condition regarding the purchase date.

[0091] As described above, the recommended product determination unit 1105 may determine whether each related product meets the recommendation criteria in relation to all the purchased products. The recommended product determination unit 1105 may then determine the related product that meets the recommendation criteria in relation to all the purchased products as the recommended product.

[0092] The recommended product determination unit 1105 may calculate a score based on similarity and the age of purchase. For example, the recommended product determination unit 1105 may calculate a score using a predetermined formula with similarity and the age of purchase as variables. In this case, the recommendation condition may be that the calculated score is above (or below) a predetermined threshold value. The calculation of the score in this case may correspond to either a correction of similarity based on the age of purchase or a correction of the age of purchase based on similarity. Therefore, this recommendation condition is essentially either a condition for similarity or a condition for the age of purchase. The recommended product determination unit 1105 may calculate a score such that, for example, the lower the similarity, the higher the score, and the older the purchase date, the higher the score.

[0093] The recommended product information transmission unit 1106 may transmit recommended product information, which indicates the recommended product determined by the recommended product determination unit 1105, to the user terminal 2 of the recommended user. The recommended product information may include at least one of the following: the product name and product image of the recommended product. The recommended product information may also include the URL of the HTML document of the recommended product's product page. This allows the recommended product information transmission unit 1106 to enable the user terminal 2 to send a request for the HTML document of the recommended product to the online shopping server 1 when the recommended user selects any of the recommended products from the recommended product information. Examples of recommended product information include HTML documents, web pages, emails, and instant messages. By presenting the recommended product information on the user terminal 2, the user can view the recommended products.

[0094] [4. Operation of the communication system] Next, the operation of the communication system S will be described with reference to Figures 9 and 10. The system control unit 11 of the online shopping server 1 may execute the processes shown in Figures 9 and 10 according to the program code included in the transaction service program. The processes shown in these figures are illustrative, and any processes may be executed as long as the objective is achieved. The order of the processes is not limited to the order shown in these figures. Also, at least one of the steps shown in these figures may be omitted.

[0095] Figure 9 is a flowchart showing an example of the top page transmission process executed by the system control unit 11 of the online shopping server 1 according to this embodiment. For example, a user who is to be recommended performs an operation to display the top page of the trading site. In response to this operation, the user terminal 2 sends a request for the top page to the online shopping server 1. This request may include the user ID of the recommended user. The online shopping server 1 may execute the top page transmission process in response to receiving this request. As shown in Figure 9, the system control unit 11 executes the recommended product determination process (step S101).

[0096] Figure 10 is a flowchart showing an example of the recommended product determination process executed by the system control unit 11 of the online shopping server 1 according to this embodiment. As shown in Figure 10, the purchase history acquisition unit 1101 searches the purchase history DB 14c for all purchase logs, including the user ID of the recommended user (step S201). The user ID of the recommended user can be obtained from a request received from the user terminal 2. The purchase history acquisition unit 1101 identifies the products indicated by the purchased product information contained in each retrieved purchase log as products purchased by the recommended user (step S202). Next, the product-related information acquisition unit 1103 acquires a vector associated with the combination of the store ID and product ID of each purchased product from the product vector DB 14d (step S203). The acquired vector may be two or all of the following: text vector, image vector, and simultaneous viewing status vector. Next, the product-related information acquisition unit 1103 integrates the acquired vectors for each purchased product to generate a vector for similarity calculation (step S204).

[0097] Next, the related product identification unit 1102 identifies related products for each product identified as a purchased product (step S205). For example, the related product identification unit 1102 may search the concurrently purchased product DB 14e for concurrently purchased product information associated with the combination of the store ID and product ID of the purchased product. The related product identification unit 1102 may identify the products indicated by the retrieved concurrently purchased product information as related products. The related product identification unit 1102 may also obtain category information of the purchased product from the product DB 14a. The related product identification unit 1102 may search the product DB 14a for product information that includes the same category information as the obtained category information. The related product identification unit 1102 may identify the products corresponding to the retrieved product information as related products. The related product identification unit 1102 may also obtain brand or series identification information of the purchased product from the catalog DB. The related product identification unit 1102 may search the catalog DB for product codes associated with the same identification information as the obtained identification information. The related product identification unit 1102 may search the product database 14a for product information containing the same product code as the retrieved product code. The related product identification unit 1102 may identify the product corresponding to the retrieved product information as a related product. The related product identification unit 1102 determines the identified related product as a candidate product for recommended products.

[0098] Next, the product relationship information acquisition unit 1103 acquires a vector from the product vector DB 14d associated with the store ID and candidate product ID of each related product identified as a candidate product (step S206). Then, for each candidate product, the product relationship information acquisition unit 1103 integrates the acquired vectors to generate a vector for similarity calculation (step S207).

[0099] Next, the recommended product determination unit 1105 selects one of the products purchased by the recommended user (step S208). The selected product is called the selected purchased product. Next, the recommended product determination unit 1105 calculates the age of the purchase date of the selected purchased product (step S209). For example, the recommended product determination unit 1105 obtains the order date and time from the purchase log of the selected product among the acquired purchase logs. The recommended product determination unit 1105 determines the age of the purchase date by calculating the length of time that has elapsed from the order date and time to today. Next, the recommended product determination unit 1105 determines a correction coefficient according to the age of the purchase date (step S210). The recommended product determination unit 1105 may make the correction coefficient smaller the older the purchase date.

[0100] Next, the recommended product determination unit 1105 selects one of the identified candidate products (step S211). Next, the similarity calculation unit 1104 calculates the similarity between the vector for calculating the similarity of the selected purchase product and the vector for calculating the similarity of the selected candidate product (step S212). Next, the similarity calculation unit 1104 calculates the corrected similarity by multiplying the calculated similarity by a correction coefficient (step S213). Next, the recommended product determination unit 1105 determines whether the corrected similarity is equal to or greater than a predetermined threshold value (step S214). If the corrected similarity is equal to or greater than the threshold value (step S214: YES), the recommended product determination unit 1105 removes the related product that was selected as a candidate product from the list of candidate products (step S215). Related products that have been removed from the list of candidate products will no longer be selected in step S211.

[0101] After step S215, or if the corrected similarity is below the threshold value (step S214: NO), the recommended product determination unit 1105 determines whether all candidate products have been selected for the selected purchase product (step S216). If there are candidate products that have not been selected (step S216: NO), the process proceeds to step S211. In step S211, the recommended product determination unit 1105 selects one candidate product from among the related products that remain as candidate products and have not yet been selected. If all candidate products have been selected (step S216: YES), the recommended product determination unit 1105 determines whether all purchased products have been selected (step S217). If there are purchased products that have not yet been selected (step S217: NO), the process proceeds to step S208. In step S208, the recommended product determination unit 1105 selects one product from among the purchased products that have not yet been selected. If all purchased items have been selected (step S217: YES), the recommended product determination unit 1105 determines the remaining related products as recommended products (step S218). After step S218 is completed, the recommended product determination process ends.

[0102] Returning to Figure 9, once the recommended product determination process is complete, the recommended product information transmission unit 1106 generates an HTML document for the top page containing information about the recommended products (step S102). For example, the recommended product information transmission unit 1106 may obtain product information for each recommended product from the product DB 14a. Based on the product information, the recommended product information transmission unit 1106 may add the product name, product image URL, price, store name, URL of the product page's HTML document, etc., to the HTML document for the top page. Next, the recommended product information transmission unit 1106 sends the generated HTML document to the user terminal 2 (step S103), and the top page transmission process ends. The user terminal 2 displays the top page based on this HTML document. The top page contains information about each recommended product.

[0103] As described above, according to this embodiment, related products that have a predetermined relationship with the product purchased by the recommended user are identified. The similarity between the product relationship information of the purchased product and the product relationship information of the related products is calculated. Then, recommended products are determined based on the purchase date and similarity of the product purchased by the recommended user. When determining recommended products, the ease of satisfying either the similarity or the age of the purchase date changes depending on the other. Therefore, by taking into account the time the user purchased the product, it is possible to recommend products that match the user's preferences.

[0104] Here, the online shopping server 1 may change the similarity by adjusting it according to the age of the purchase date, or it may change the baseline value according to the age of the purchase date. In this case, by adjusting the similarity according to the age of the purchase date, or by changing the baseline value according to the age of the purchase date, the degree to which the conditions for similarity are easily satisfied can be changed according to the age of the purchase date.

[0105] In this case, the online shopping server 1 may adjust the similarity or determine a baseline value so that the older the purchase date, the more likely the comparison result is to reach a predetermined outcome. In this case, the older the purchase date, the more likely the comparison result is to reach a predetermined outcome. Therefore, the older the purchase date, the easier it is to satisfy the conditions for similarity. Consequently, the older the purchase date of a product purchased by a user, the higher the probability that the related product will become a recommended product.

[0106] Furthermore, the online shopping server 1 may adjust the similarity so that the adjusted similarity becomes lower the older the purchase date, or it may determine a higher baseline value. In this case, the older the purchase date, the easier it is to satisfy the conditions for similarity, and the lower the similarity, the higher the probability of satisfying those conditions. In this case, related products similar to products purchased relatively recently will be less likely to satisfy the conditions for similarity, and therefore will have a relatively low probability of becoming recommended products. Related products similar to products purchased relatively old will be more likely to satisfy the conditions for similarity, and therefore will have a relatively high probability of becoming recommended products. Even if a user has purchased a product relatively recently, related products that are not similar to that product will have a higher probability of becoming recommended products than similar related products. Therefore, by considering the timing of the user's purchase, it is possible to recommend products that match the user's preferences.

[0107] Alternatively, the online shopping server 1 may determine a correction factor for similarity according to the age of the purchase, and calculate the corrected similarity based on the determined correction factor and the similarity. In this case, by determining a correction factor for similarity according to the age of the purchase, the degree to which the conditions for similarity are easily satisfied can be changed.

[0108] Furthermore, the online shopping server 1 may increase the ease of satisfying the similarity criteria as the purchase date increases. In this case, the older the purchase date of a product purchased by a user, the higher the probability that related products will become recommended products.

[0109] Furthermore, the online shopping server 1 may calculate a first similarity between the product relationship information of the related product and the product relationship information of any first product among the purchased products. Also, if the online shopping server 1 determines that the related product satisfies the conditions for a recommended product based on the first similarity and the age of the first product's purchase, it may calculate a second similarity between the product relationship information of the related product and the product relationship information of any second product among the purchased products. Furthermore, if the online shopping server 1 determines, by the determination means, that the related product does not satisfy the conditions for a recommended product, it does not need to calculate the second similarity. In this case, the similarity calculation that would be unnecessary when calculating similarity between the related product and each of the purchased products to determine a recommended product can be omitted.

[0110] Here, the online shopping server 1 may designate a related product as a recommended product if it determines that all of the related products meet the criteria for a recommended product for all of the purchased products. In this case, it is possible to determine a product that is suitable to be a recommended product for any of the products purchased by the user.

[0111] Furthermore, the online shopping server 1 may calculate similarity based on text vectors and image vectors. Alternatively, the online shopping server 1 may calculate similarity based on browsing status vectors and text vectors. In these cases, the similarity between purchased items and related items can be determined more appropriately.

[0112] Here, the online shopping server 1 may integrate the vectors included in the product relationship information for each purchased product and related product to generate a vector for similarity calculation, and then calculate the similarity of the vector for similarity calculation. In this case, by integrating the vectors, the similarity can be calculated in a single calculation.

[0113] Furthermore, the online shopping server 1 may identify products purchased by the same user at the same time as the product purchased by the recommended user, based on purchase history information showing the purchase history of products by multiple users. In this case, products purchased by the same user at the same time as the product purchased by the recommended user are included in the related products. The recommended user may also want products purchased at the same time as the product purchased by the recommended user. Products purchased at the same time may or may not be similar to each other. Therefore, related products may include products that are relatively dissimilar to the purchased product, as products that the recommended user may want.

[0114] Furthermore, the online shopping server 1 may increase the ease with which the condition regarding the age of the purchase is met as the similarity is lower. In this case, the lower the similarity, the higher the probability that the related product will become a recommended product.

[0115] Here, the online shopping server 1 may adjust the purchase date so that the lower the similarity, the older the purchase date, or it may determine the age of the standard so that the standard is newer. In this case, the lower the similarity, the easier it is to satisfy the condition for similarity, and the older the purchase date, the higher the probability of satisfying that condition.

[0116] (Note 1) Purchase history information showing the purchase history of products by the target user, comprising: purchase history information acquisition means for acquiring purchase history information showing the purchased product and the time when the product was purchased; identification means for identifying related products that have a predetermined relationship with the purchased product shown by the acquired purchase history information; product relationship information acquisition means for acquiring product relationship information relating to a product, comprising: product relationship information acquisition means for acquiring the product relationship information of the purchased product and the product relationship information of the identified related product; and the product relationship information of the purchased product and the product relationship information of the identified related product A recommended product determination device comprising: a calculation means for calculating similarity with related information; a determination means for determining a recommended product from among the identified related products that satisfies the conditions for a product to be recommended to the user, based on the age of the purchase date of the purchased product and the calculated similarity; and a recommended product information transmission means for transmitting recommended product information indicating the determined recommended product to the user's terminal device, wherein the conditions for the recommended product include a condition for either the similarity or the age of the purchase date, and the determination means changes the degree of ease of satisfying the condition for one of them according to the other.

[0117] (Note 2) The recommended product determination device according to Note 1, wherein the conditions for the recommended product include the result of comparing the similarity with a reference value for the similarity, and the determination means changes the similarity by making a correction to the calculated similarity according to the age of the purchase date, or changes the reference value according to the age of the purchase date.

[0118] (Note 3) The recommended product determination device according to Note 2, characterized in that the determination means corrects the similarity or determines the reference value such that the older the purchase date, the more likely the comparison result is to be the predetermined result.

[0119] (Note 4) The recommended product determination device according to Note 3, wherein the conditions for the recommended product include the similarity being less than the standard value, and the determination means corrects the similarity so that the corrected similarity becomes lower the older the purchase date, or determines the standard value so that the standard value becomes higher.

[0120] (Note 5) The recommended product determination device according to any one of Notes 2 to 4, characterized in that the determination means determines a correction coefficient for the similarity according to the age of the purchase date, and calculates the corrected similarity based on the determined correction coefficient and the similarity.

[0121] (Note 6) The recommended product determination device according to any one of Notes 1 to 5, wherein the conditions for the recommended product include the conditions for similarity, and the determination means increases the degree of ease of satisfying the conditions for similarity as the purchase date is older.

[0122] (Note 7) The recommended product determination device according to any one of Notes 1 to 6, wherein the purchased product includes the first product and the second product, the calculation means calculates a first similarity between the product relationship information of the related product and the product relationship information of the first product, and if the determination means determines that the related product satisfies the conditions for the recommended product based on the first similarity and the age of the purchase date of the first product, it calculates a second similarity between the product relationship information of the related product and the product relationship information of the second product, and if the determination means determines that the related product does not satisfy the conditions for the recommended product based on the first similarity and the age of the purchase date of the first product, it does not calculate the second similarity.

[0123] (Note 8) The recommended product determination device according to Note 7, characterized in that the determination means determines the related products that meet the conditions for recommended products for all of the related products that have been purchased as the recommended products.

[0124] (Note 9) The recommended product determination device according to any one of Notes 1 to 8, wherein the acquired product-related information includes a text vector and an image vector of the product posted on a website of a trading service that enables the purchase of the product corresponding to the product-related information, and the calculation means calculates the similarity based on the text vector and the image vector.

[0125] (Note 10) The recommended product determination device according to any one of Notes 1 to 9, wherein the acquired product-related information includes a vector indicating whether the same user viewed each of multiple products on a website of a trading service that enables the purchase of products and the product to which the product-related information corresponds at the same time, and a vector of text posted on the website for the product to which the product-related information corresponds, and the calculation means calculates the similarity based on the vector indicating the viewing status and the vector of the text.

[0126] (Note 11) The recommended product determination device according to any one of Notes 1 to 10, wherein the acquired product-related information includes a vector indicating whether the same user viewed each of multiple products on a website of a trading service that enables the purchase of products and the product to which the product-related information corresponds at the same time, and a vector of an image of the product to which the product-related information corresponds, and the calculation means calculates the similarity based on the vector indicating the viewing status and the image vector.

[0127] (Note 12) The recommended product determination device according to any one of Notes 9 to 11, characterized in that the calculation means integrates the vectors included in the acquired product relationship information for each of the purchased product and the related product to generate a vector for similarity calculation, and calculates the similarity of the vector for similarity calculation.

[0128] (Note 13) The recommended product determination device according to any one of Notes 1 to 12, wherein the products having the predetermined relationship include products purchased by the same user at the same time as the products indicated by the acquired purchase history information, and the identifying means identifies the products having the predetermined relationship based on second purchase history information indicating the purchase history of products by multiple users, the products purchased, the time when the products were purchased, and the user who purchased the products.

[0129] (Note 14) The recommended product determination device according to any one of Notes 1, 7 to 13, characterized in that the conditions for the recommended product include the condition regarding the age of the purchase date, and the determination means increases the degree to which the condition regarding the age of the purchase date is easily satisfied as the similarity decreases.

[0130] (Note 15) The recommended product determination device according to Note 14, wherein the condition for the recommended product is that the purchase date is older than the standard age, and the determination means corrects the purchase date age so that the lower the similarity, the older the purchase date becomes, or determines the standard age so that the standard age becomes newer.

[0131] (Note 16) In a recommended product determination method performed by a computer, the method includes: a purchase history information acquisition step that acquires purchase history information indicating the purchase history of products by a user of a terminal device, which includes the purchased product and the time when the product was purchased; a identification step that identifies a related product that has a predetermined relationship with the purchased product indicated by the acquired purchase history information; a product relationship information acquisition step that acquires product relationship information relating to a product, which includes the product relationship information of the purchased product and the product relationship information of the identified related product; and the product relationship information of the purchased product A method for determining a recommended product, comprising: a calculation step of calculating the similarity between information and the product relationship information of the identified related product; a determination step of determining a recommended product from among the identified related products that satisfies the conditions for a product to be recommended to the user, based on the age of the purchase date of the purchased product and the calculated similarity; and a recommended product information transmission step of transmitting recommended product information indicating the determined product to the terminal device, wherein the conditions for the recommended product include a condition for either the similarity or the age of the purchase date, and the determination step changes the degree of ease of satisfying the condition for one of them according to the other.

[0132] (Note 17) A computer includes purchase history information acquisition means that acquires purchase history information showing the purchase history of goods by a user of a terminal device, which indicates the purchased goods and the time when the goods were purchased; identification means that identify related goods that have a predetermined relationship with the purchased goods indicated by the acquired purchase history information; product relationship information acquisition means that acquires product relationship information relating to goods, which acquires the product relationship information of the purchased goods and the product relationship information of the identified related goods; and product relationship information acquisition means that acquires the product relationship information of the purchased goods and the identified related goods A recommended product determination program comprising: a calculation means for calculating the similarity of an item to the aforementioned product-related information; a determination means for determining a recommended product from among the identified related products that satisfies the conditions for a product to be recommended to the user, based on the age of the purchase date of the purchased product and the calculated similarity; and a recommended product information transmission means for transmitting recommended product information indicating the determined product to the terminal device, wherein the conditions for the recommended product include a condition for either the similarity or the age of the purchase date, and the determination means changes the degree of ease of satisfying the condition for one of the two according to the other. [Explanation of Symbols]

[0133] 1. Online Shopping Server 2 User terminals 11 System Control Unit 12 System Bus 13 Input / Output Interfaces 14 Storage section 15 Communications Department 14a Product DB 14b Browsing History Database 14c Purchase History Database 14d Product Vector Database 14e DB of products purchased at the same time 1101 Purchase History Acquisition Section 1102 Related Products Identification Department 1103 Product-Related Information Acquisition Department 1104 Similarity calculation part 1105 Recommended Product Selection Department 1106 Recommended Product Information Transmission Department NW Network S Communication System

Claims

1. Purchase history information that shows the purchase history of a target user, and purchase history information that shows the purchased product and the time when the product was purchased, A means for identifying related products that have a predetermined relationship with the purchased product indicated by the acquired purchase history information, Product-related information acquisition means for acquiring product-related information relating to a product, comprising product-related information acquisition means for acquiring the product-related information of the purchased product and the product-related information of the specified related product, A calculation means for calculating the similarity between the product-related information of the purchased product and the product-related information of the identified related product, A determination means for determining a recommended product from among the identified related products that satisfies the conditions for a product to be recommended to the user, based on the age of the purchase date of the purchased product and the calculated similarity; A recommended product information transmission means transmits recommended product information indicating the determined recommended product to the user's terminal device. Equipped with, The aforementioned conditions for the recommended products include conditions for either the similarity or the age of the purchase date. The aforementioned determination means is a recommended product determination device characterized by changing the degree of ease of satisfying one of the conditions according to the other.

2. The conditions for the recommended product include that the result of comparing the similarity with a baseline value for that similarity is a predetermined result, The recommended product determination device according to claim 1, characterized in that the determination means changes the similarity by correcting the calculated similarity according to the age of the purchase date, or changes the reference value according to the age of the purchase date.

3. The recommended product determination device according to claim 2, characterized in that the determination means corrects the similarity or determines the reference value such that the older the purchase date, the more likely the comparison result is to be the predetermined result.

4. The conditions for the recommended products include the similarity being less than the threshold value, The recommended product determination device according to claim 3, characterized in that the determination means corrects the similarity so that the corrected similarity becomes lower the older the purchase date, or determines the reference value so that the reference value becomes higher.

5. The recommended product determination device according to any one of claims 2 to 4, characterized in that the determination means determines a correction coefficient for the similarity according to the age of the purchase date, and calculates the corrected similarity based on the determined correction coefficient and the similarity.

6. The conditions for the aforementioned recommended products include the conditions for similarity, The recommended product determination device according to any one of claims 1 to 4, characterized in that the determination means increases the degree of ease of satisfying the conditions for similarity as the purchase date is older.

7. The purchased goods include the first and second goods, The recommended product determination device according to any one of claims 1 to 4, characterized in that the calculation means calculates a first similarity between the product relationship information of the related product and the product relationship information of the first product, and if the determination means determines that the related product satisfies the conditions for the recommended product based on the first similarity and the age of the purchase date of the first product, it calculates a second similarity between the product relationship information of the related product and the product relationship information of the second product, and if the determination means determines that the related product does not satisfy the conditions for the recommended product based on the first similarity and the age of the purchase date of the first product, it does not calculate the second similarity.

8. The recommended product determination device according to claim 7, characterized in that the determination means determines the related products that meet the conditions for recommended products for all of the related products that have been purchased as the recommended products.

9. The acquired product-related information includes, for the product to which the product-related information corresponds, a vector of text and a vector of an image of the product that are posted on the website of the trading service that enables the purchase of the product. The recommended product determination device according to claim 1, characterized in that the calculation means calculates the similarity based on the vector of the text and the vector of the image.

10. The acquired product-related information includes a vector indicating whether the same user viewed each of multiple products on a website of a trading service that enables the purchase of products and the product to which the product-related information corresponds at the same time, and a vector of text posted on the website for the product to which the product-related information corresponds. The recommended product determination device according to claim 1, characterized in that the calculation means calculates the similarity based on the vector indicating the browsing status and the vector of the text.

11. The acquired product-related information includes a vector indicating whether the same user viewed each of multiple products on a website of a trading service that enables the purchase of products and the product to which the product-related information corresponds at the same time, and a vector of an image of the product to which the product-related information corresponds. The recommended product determination device according to claim 1, characterized in that the calculation means calculates the similarity based on the vector indicating the browsing status and the vector of the image.

12. The recommended product determination device according to any one of claims 9 to 11, characterized in that the calculation means integrates the vectors included in the acquired product relationship information for each of the purchased product and the related product to generate a vector for similarity calculation, and calculates the similarity of the vector for similarity calculation.

13. The products having the aforementioned predetermined relationship include products purchased by the same user at the same time as the products indicated by the acquired purchase history information. The recommended product determination device according to claim 1, characterized in that the identifying means is second purchase history information showing the purchase history of products by multiple users, and the device identifies products having the predetermined relationship based on the second purchase history information showing the purchased product, the time when the product was purchased, and the user who purchased the product.

14. The conditions for the aforementioned recommended products include the condition regarding the age of the purchase date, The recommended product determination device according to claim 1, characterized in that the determination means increases the degree to which the condition regarding the age of the purchase date is easily satisfied as the similarity decreases.

15. The condition for the aforementioned recommended product is that the purchase date is older than the standard age. The recommended product determination device according to claim 14, characterized in that the determination means corrects the age of the purchase date so that the lower the similarity, the older the purchase date becomes, or determines the age of the standard so that the standard becomes newer.

16. In a computer-based method for determining recommended products, Purchase history information that shows the purchase history of products by a user of a terminal device, a purchase history information acquisition step that acquires purchase history information indicating the purchased products and the time when the products were purchased, A selection step to identify related products that have a predetermined relationship with the purchased product indicated by the acquired purchase history information, A product-related information acquisition step for acquiring product-related information relating to a product, comprising: a product-related information acquisition step for acquiring the product-related information of the purchased product and the product-related information of the identified related product; A calculation step of calculating the similarity between the product-related information of the purchased product and the product-related information of the identified related product, A decision step in which, based on the purchase date of the purchased product and the calculated similarity, a recommended product is determined from among the identified related products that satisfies the conditions for a product to be recommended to the user, A recommended product information transmission step involves transmitting recommended product information indicating the determined product to the terminal device. Includes, The aforementioned conditions for the recommended products include conditions for either the similarity or the age of the purchase date. The method for determining recommended products is characterized in that the determination step involves changing the degree of ease of satisfying one of the conditions according to the other.

17. Computers, Purchase history information that shows the purchase history of products by a user of a terminal device, and purchase history information acquisition means that acquires purchase history information indicating the purchased products and the time when the products were purchased, A means for identifying related products that have a predetermined relationship with the purchased product indicated by the acquired purchase history information, Product-related information acquisition means for acquiring product-related information relating to a product, comprising product-related information acquisition means for acquiring the product-related information of the purchased product and the product-related information of the specified related product, A calculation means for calculating the similarity between the product-related information of the purchased product and the product-related information of the identified related product, A determination means for determining a recommended product from among the identified related products that satisfies the conditions for a product to be recommended to the user, based on the age of the purchase date of the purchased product and the calculated similarity; Recommended product information transmission means for transmitting recommended product information indicating the determined product to the terminal device, To make it function as, The aforementioned conditions for the recommended products include conditions for either the similarity or the age of the purchase date. The aforementioned determination means is a recommended product determination program characterized by changing the degree of ease of satisfying one of the conditions according to the other.

Images