CN118115240A

CN118115240A - Product recommendation method, model training method, device, equipment, medium and product

Info

Publication number: CN118115240A
Application number: CN202410285308.9A
Authority: CN
Inventors: 贺天生; 张嘏晖; 支一飞; 张磊
Original assignee: Industrial and Commercial Bank of China Ltd ICBC
Current assignee: Industrial and Commercial Bank of China Ltd ICBC
Priority date: 2024-03-13
Filing date: 2024-03-13
Publication date: 2024-05-31

Abstract

The present disclosure provides a product recommendation method, apparatus, device, storage medium, and program product, which can be applied to the technical field of artificial intelligence and the technical field of financial science and technology. The product recommendation method comprises the following steps: obtaining browsing characteristics of a target user based on a browsing record set of the target user; generating product features based on the product data graph model; and inputting the browsing characteristics and the product characteristics into a product recommendation network to generate a product recommendation result for the target user, wherein the product data graph model comprises a plurality of product nodes and a plurality of edges, the node information of each product node comprises product attribute information, and the edge information of each edge represents a product association relationship between two products corresponding to two product nodes connected with the edge. The disclosure also provides a product recommendation network training method and device.

Description

Product recommendation method, model training method, device, equipment, medium and product

Technical Field

The present disclosure relates to the field of artificial intelligence technology and the field of financial technology, and more particularly, to a product recommendation method, a model training method, an apparatus, a device, a medium, and a program product.

Background

With the development of artificial intelligence technology, recommendation models constructed based on the artificial intelligence technology are widely applied to product recommendation systems.

In the prior art, product recommendation is typically implemented based on collaborative filtering, considering popular products or products associated with or similar to popular products for recommendation to users. But such recommendations are typically focused on hot products and ignore newly released products or cold products, resulting in product recommendations that do not match the user's expectations.

Disclosure of Invention

In view of the foregoing, the present disclosure provides a product recommendation method, a model training method, an apparatus, a device, a medium, and a program product.

According to a first aspect of the present disclosure, there is provided a product recommendation method, comprising: obtaining browsing characteristics of the target user based on the browsing record set of the target user; generating product features based on the product data graph model; and inputting the browsing features and the product features into a product recommendation network to generate a product recommendation result for the target user, wherein the product data graph model comprises a plurality of product nodes and a plurality of edges, node information of each of the plurality of product nodes comprises product attribute information, and edge information of each of the plurality of edges characterizes a product association relationship between two products corresponding to two product nodes connected with the edge.

According to an embodiment of the present disclosure, the product recommendation network includes at least one stacked picture scroll layer and a mapping layer, and the inputting the browsing feature and the product feature into the product recommendation network generates a product recommendation result for the target user, including: inputting the browsing feature and the product feature into the at least one picture scroll layer to obtain a convolution feature; inputting the convolution characteristics into the mapping layer, and determining the purchase probability of each product related to the target user purchasing the product data graph model; and determining the product recommendation result based on the purchase probabilities of the products.

According to an embodiment of the present disclosure, the generating, based on the browsing record set of the target user, browsing features of the target user includes: determining initial browsing characteristics of the target user based on the browsing record set; and inputting the initial browsing characteristics into a preset three-dimensional convolution network, and carrying out characteristic enhancement on the initial browsing characteristics to generate the browsing characteristics.

According to an embodiment of the present disclosure, the initial browsing feature includes a first sub-feature, a second sub-feature, and a third sub-feature, and the determining the initial browsing feature of the target user based on the browsing record set includes: determining a first data set from the browsing record set, and performing vectorization representation to obtain the first sub-feature, wherein the browsing amount of the products in the first data set in a first history period is higher than the browsing amount of other products in the browsing record in the first history period; determining a second data set from the browse record set, and performing vectorization representation to obtain the second sub-feature, wherein the browse amount of the products in the second product set in a second history period is higher than the browse amount of other products in the browse record in the second history period, and the second history period is longer than the first history period; determining the respective browsing times of each product category from the browsing record set, and performing vectorization representation to obtain the third sub-feature; and obtaining the browsing data feature based on the first sub-feature, the second sub-feature and the third sub-feature.

According to an embodiment of the present disclosure, the generating product features based on the product data graph model includes: generating an adjacency matrix of the product data graph model; obtaining a degree matrix of the product data graph model based on the adjacent matrix; and obtaining the product characteristics based on the adjacency matrix and the degree matrix.

According to an embodiment of the present disclosure, the determining, from the browsing record set, a respective browsing number of times of each product category, and performing vectorized representation, to obtain the third sub-feature, includes: determining the browsing times of the target user for each product category based on the browsing record set; vectorizing the browsing times according to the preset arrangement sequence of each product category to generate an initial third sub-feature; and normalizing the initial third sub-feature to obtain the third sub-feature.

A second aspect of the present disclosure provides a product recommendation network training comprising: based on a sample browsing record set of a sample user, obtaining sample browsing characteristics of the sample user; generating sample product features based on the sample product data graph model; inputting the sample browsing characteristics and the sample product characteristics into a product recommendation network to generate a sample product recommendation result for the sample user, wherein the product data graph model comprises a plurality of product nodes and a plurality of edges, node information of each of the plurality of product nodes comprises product attribute information, and edge information of each of the plurality of edges represents a product association relationship between two products corresponding to two product nodes connected with the edge; and training the product recommendation network based on the sample product recommendation result and a label matched with the sample browsing record set, wherein the label comprises the interesting product result of the sample user.

A third aspect of the present disclosure provides a product recommendation device, comprising: the browsing characteristic generation module is used for obtaining the browsing characteristics of the target user based on the browsing record set of the target user; the product characteristic generating module is used for generating product characteristics based on the product data graph model; and a recommendation result generating module, configured to input the browsing feature and the product feature into a product recommendation network, and generate a product recommendation result for the target user, where the product recommendation network is obtained by training a graph convolutional neural network using a product data graph model, the product data graph model includes a plurality of nodes and a plurality of edges, each of the nodes corresponds to a product, node information of each of the plurality of nodes includes a product identifier, and an edge information representation of each of the plurality of edges has an association relationship with two nodes corresponding to the edge information.

A fourth aspect of the present disclosure provides a product recommendation network training apparatus, comprising: the sample browsing feature generation module is used for obtaining sample browsing features of the sample user based on a sample browsing record set of the sample user; the sample product feature generation module is used for generating sample product features based on the sample product data graph model; the sample recommendation result generation module is used for inputting the sample browsing characteristics and the sample product characteristics into a product recommendation network to generate sample product recommendation results for the sample users, wherein the product data graph model comprises a plurality of product nodes and a plurality of edges, node information of each of the plurality of product nodes comprises product attribute information, and side information of each of the plurality of edges represents a product association relationship between two products corresponding to two product nodes connected with the edge; and a model training module, configured to train the product recommendation network based on the sample product recommendation result and a label matched with the sample browsing record set, where the label includes a product result of interest of the sample user.

A fifth aspect of the present disclosure provides an electronic device, comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method.

The sixth aspect of the present disclosure also provides a computer readable storage medium having stored thereon a computer program or instructions which when executed by a processor, perform the steps of the above method.

A seventh aspect of the present disclosure also provides a computer program product comprising a computer program or instructions which, when executed by a processor, performs the steps of the method described above.

According to the embodiment of the disclosure, the product characteristics are generated by utilizing the product graph model, so that the product characteristics comprise characteristic information of the whole quantity of products, the similarity among the inherent products is more focused, the product recommendation result is prevented from being concentrated on hot products and the cold products are prevented from being ignored, and the obtained product recommendation result is more accurate.

Drawings

The foregoing and other objects, features and advantages of the disclosure will be more apparent from the following description of embodiments of the disclosure with reference to the accompanying drawings, in which:

FIG. 1 schematically illustrates an application scenario diagram of a product recommendation method, apparatus, device, medium and program product according to an embodiment of the present disclosure;

FIG. 2 schematically illustrates a flow chart of a product recommendation method according to an embodiment of the present disclosure;

FIG. 3 schematically illustrates a schematic diagram of a product data map model in accordance with a specific embodiment of the present disclosure;

FIG. 4 schematically illustrates a data flow diagram of a product recommendation method according to an embodiment of the present disclosure;

FIG. 5 schematically illustrates a schematic diagram of a three-dimensional convolution network according to a specific embodiment of this disclosure;

FIG. 6 schematically illustrates a schematic diagram of stacked multiple graph convolutional layers, in accordance with an embodiment of the present disclosure;

FIG. 7 schematically illustrates a flow chart of a product recommendation network training method according to an embodiment of the present disclosure;

FIG. 8 schematically illustrates a block diagram of a product recommendation device, according to an embodiment of the present disclosure;

FIG. 9 schematically illustrates a block diagram of a product recommendation network training device in accordance with an embodiment of the present disclosure; and

Fig. 10 schematically illustrates a block diagram of an electronic device adapted to implement a product recommendation method, according to an embodiment of the disclosure.

Detailed Description

Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. It should be understood that the description is only exemplary and is not intended to limit the scope of the present disclosure. In the following detailed description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the embodiments of the present disclosure. It may be evident, however, that one or more embodiments may be practiced without these specific details. In addition, in the following description, descriptions of well-known structures and techniques are omitted so as not to unnecessarily obscure the concepts of the present disclosure.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The terms "comprises," "comprising," and/or the like, as used herein, specify the presence of stated features, steps, operations, and/or components, but do not preclude the presence or addition of one or more other features, steps, operations, or components.

All terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art unless otherwise defined. It should be noted that the terms used herein should be construed to have meanings consistent with the context of the present specification and should not be construed in an idealized or overly formal manner.

Where a convention analogous to "at least one of A, B and C, etc." is used, in general such a convention should be interpreted in accordance with the meaning of one of skill in the art having generally understood the convention (e.g., "a system having at least one of A, B and C" would include, but not be limited to, systems having a alone, B alone, C alone, a and B together, a and C together, B and C together, and/or A, B, C together, etc.).

In the technical solution of the present disclosure, the related user information (including, but not limited to, user personal information, user image information, user equipment information, such as location information, etc.) and data (including, but not limited to, data for analysis, stored data, displayed data, etc.) are information and data authorized by the user or sufficiently authorized by each party, and the related data is collected, stored, used, processed, transmitted, provided, disclosed, applied, etc. in compliance with relevant laws and regulations and standards, necessary security measures are taken, no prejudice to the public order colloquia is provided, and corresponding operation entries are provided for the user to select authorization or rejection.

In the scenario of using personal information to make an automated decision, the method, the device and the system provided by the embodiment of the disclosure provide corresponding operation inlets for users, so that the users can choose to agree or reject the automated decision result; if the user selects refusal, the expert decision flow is entered. The expression "automated decision" here refers to an activity of automatically analyzing, assessing the behavioral habits, hobbies or economic, health, credit status of an individual, etc. by means of a computer program, and making a decision. The expression "expert decision" here refers to an activity of making a decision by a person who is specializing in a certain field of work, has specialized experience, knowledge and skills and reaches a certain level of expertise.

Product recommendation achieved through collaborative filtering easily results in a centralized recommendation result for hot products, and cold products are ignored, so that the recommendation result is inaccurate. And when the behavior data of the user is sparse, a proper recommended product is difficult to obtain by utilizing collaborative filtering, and the recommendation effect is poor.

The embodiment of the disclosure provides a product recommendation method, which is based on a browsing record set of a target user to obtain browsing characteristics of the target user; generating product features based on the product data graph model; and inputting the browsing characteristics and the product characteristics into a product recommendation network to generate a product recommendation result for the target user, wherein the product data graph model comprises a plurality of product nodes and a plurality of edges, the node information of each product node comprises product attribute information, and the edge information of each edge represents a product association relationship between two products corresponding to two product nodes connected with the edge.

Fig. 1 schematically illustrates an application scenario diagram of a product recommendation method according to an embodiment of the present disclosure.

As shown in fig. 1, an application scenario 100 according to this embodiment may include terminal devices 101, 102, 104, a network 104, and a server 105. The network 104 is used as a medium to provide communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, among others.

The user may interact with the server 105 via the network 104 using the terminal devices 101, 102, 103 to receive or send messages or the like. Various communication client applications, such as shopping class applications, web browser applications, search class applications, instant messaging tools, mailbox clients, social platform software, etc. (by way of example only) may be installed on the terminal devices 101, 102, 103.

The terminal devices 101, 102, 103 may be a variety of electronic devices having a display screen and supporting web browsing, including but not limited to smartphones, tablets, laptop and desktop computers, and the like.

The server 105 may be a server providing various services, such as a background management server (by way of example only) providing support for websites browsed by users using the terminal devices 101, 102, 103. The background management server may analyze and process the received data such as the user request, and feed back the processing result (e.g., the web page, information, or data obtained or generated according to the user request) to the terminal device.

It should be noted that, the product recommendation method provided by the embodiments of the present disclosure may be generally performed by the server 105. Accordingly, the product recommendation device provided by the embodiments of the present disclosure may be generally disposed in the server 105. The product recommendation method provided by the embodiments of the present disclosure may also be performed by a server or a server cluster that is different from the server 105 and is capable of communicating with the terminal devices 101, 102, 103 and/or the server 105. Accordingly, the product recommendation apparatus provided by the embodiments of the present disclosure may also be provided in a server or a server cluster that is different from the server 105 and is capable of communicating with the terminal devices 101, 102, 103 and/or the server 105.

It should be understood that the number of terminal devices, networks and servers in fig. 1 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.

The product recommendation method of the disclosed embodiment will be described in detail below with reference to fig. 2 to 7 based on the scenario described in fig. 1.

Fig. 2 schematically illustrates a flow chart of a product recommendation method according to an embodiment of the present disclosure.

As shown in fig. 2, the product recommendation method of this embodiment includes operations S210 to S230.

In operation S210, browsing characteristics of the target user are obtained based on the browsing record set of the target user.

In embodiments of the present disclosure, the consent or authorization of the target user may be obtained prior to obtaining the target user's browsing record set. For example, before operation S210, a request to acquire a browsing record set may be issued to the target user. In case that the target user agrees or authorizes that the target browsing record set can be acquired, operation S210 is performed.

In the embodiment of the disclosure, a corresponding operation entry can be provided for the target user, so that the target user can choose to agree or reject the automated decision result. That is, before the product recommendation is performed according to the browsing record set of the target user, an instruction of approval or rejection of the product recommendation input by the target user through the corresponding operation portal may be obtained. And if the target user agrees to carry out product recommendation, carrying out product recommendation on the user information.

According to embodiments of the present disclosure, the target user may be a target user of a product recommendation. Before obtaining the browsing characteristics of the target user based on the browsing record set of the target user, the method may further include: responding to a target user login product display page, and acquiring a user identification of the target user; based on the user identification, a set of browsing records for the target user is determined.

According to an embodiment of the present disclosure, the browsing record set may be a browsing record set of a target user for a product, and the browsing record set may include a plurality of browsing records, and each browsing record may include a browsing time, a browsing product, product attribute information of the browsing product, and the like.

According to the embodiment of the disclosure, the browsing operation of the target user can be recorded by exposing buried points and the like, and the browsing record matched with the browsing operation is generated.

According to an embodiment of the present disclosure, the browse feature is derived from browsing the record set. Specifically, the browsing characteristics of the target user can be obtained by performing operations such as screening, vectorization and the like on the data in the browsing record set.

In operation S220, product features are generated based on the product data graph model.

According to the embodiment of the disclosure, the product data graph model can be a two-dimensional graph model, and can be generated according to actual business requirements by utilizing product attribute information. The product data graph model can comprise the total quantity of products in the business, and the cold products are required to be covered besides the hot products, so that the phenomenon that the cold products are ignored due to the fact that product recommendation is concentrated on the hot products is avoided, and the accuracy of product recommendation is improved.

According to an embodiment of the disclosure, a product data graph model includes a plurality of product nodes and a plurality of edges, node information of each of the plurality of product nodes includes product attribute information, and edge information of each of the plurality of edges characterizes a product association relationship between two products corresponding to two product nodes connected to the edge.

Product attribute information according to embodiments of the present disclosure may be inherent to the product itself, and may include product identification, product release time, product category, product price, and the like. The product association relationship can then represent that the product attribute information between two products is similar, such as similar product release time, similar product price, or similar product category.

According to the embodiment of the disclosure, the product characteristics can be obtained by extracting the characteristics of the product data graph model. Specifically, the product attribute information and the product association information may be extracted.

In operation S230, the browsing characteristics and the product characteristics are input into the product recommendation network, and a product recommendation result for the target user is generated.

According to embodiments of the present disclosure, the product recommendation network may be trained from machine learning models, deep learning networks, and the like, e.g., the product recommendation network may be trained from a graph convolution neural network. The product recommendation results may include the product of interest to the target user.

According to the embodiment of the disclosure, the product recommendation network can obtain the similarity between products by learning the characteristics of the products. And determining the interested products and the attribute information of the products of the target user by utilizing the browsing characteristics, and determining similar products based on the products and the attribute information of the products to obtain a product recommendation result.

According to an embodiment of the present disclosure, after inputting the browsing feature and the product feature into the product recommendation network, generating the product recommendation result for the target user, further includes: and generating a product display page based on the product recommendation result.

According to an embodiment of the present disclosure, the product recommendation method further includes: a product data graph model is generated. The method for generating the product data graph model comprises the following steps: attribute information of the product is determined, including a category of the product, an amount of product resources, and the like. Based on the attribute information, an association relationship between a plurality of product nodes is determined.

Fig. 3 schematically illustrates a schematic diagram of a product data map model according to a specific embodiment of the present disclosure.

As shown in FIG. 3, the product data graph model includes two dimensions of release time and resource amount of the product.

According to embodiments of the present disclosure, in the release time dimension, the products are arranged in a order of release time from first to last, e.g., release time of product 4 is earlier than release time of product 8. In the resource amount dimension, the products are arranged in order of the resource amount from small to large, for example, the resource amount of the product 4 is larger than the resource amount of the product 5.

According to the embodiment of the disclosure, every two adjacent product nodes are connected by an edge, and the edge represents that the resource amounts of two products are the same or similar and the release time is the same or similar. For example, the amount of resources for product 4 is the same as that for product 8, and the release times are similar. The release time of the product 4 is the same as that of the product 5, and the resource amount is similar. The product 5 has similar resource amount and release time as the product 8.

According to the embodiment of the disclosure, different weights can be allocated to different edges according to the requirements of users. Specifically, in the case where the user is more concerned about the amount of resources, the similarity between two products of the same amount of resources can be improved by assigning a larger weight to the edge between the two products of the same amount of resources.

According to an embodiment of the present disclosure, generating a browsing feature of a target user based on a browsing record set of the target user includes: determining initial browsing characteristics of a target user based on the browsing record set; and inputting the initial browsing characteristics into a preset three-dimensional convolution network, and carrying out characteristic enhancement on the initial browsing characteristics to generate browsing characteristics.

According to embodiments of the present disclosure, the initial browsing features may be obtained by screening and vectorizing a browsing record set. In the case where the browsing records of the target user are small, the initial browsing feature dimension of the target user is low, and the preference of the target user cannot be accurately determined. Moreover, when the number of products is large, the interaction behavior between the target user and the products is sparse. Therefore, the initial browsing feature can be enhanced, the dimension of the initial browsing feature is increased, and the browsing feature is obtained, so that the browsing feature comprises richer information.

In accordance with an embodiment of the present disclosure, before determining the initial browsing characteristics of the target user based on the browsing record set, it may further include: and cleaning the data of the browsing record set.

According to embodiments of the present disclosure, the initial browsing features may be enhanced by a three-dimensional convolutional network. When the initial browsing characteristics are enhanced through the three-dimensional convolution network, the three-dimensional convolution kernel can be split into a two-dimensional convolution kernel and a one-dimensional convolution kernel, and the overfitting can be avoided while parameters required by the three-dimensional convolution network are reduced. Specifically, the above procedure is shown in the following formula (1):

wherein, Normalized input data for layer I,/>And/>Respectively representing a j-th polarized one-dimensional kernel, a spatial two-dimensional kernel and a deviation. /(I)5×5 Sampling network for defining criteria, p _n for enumeration/>Is provided.

According to embodiments of the present disclosure, the number of parameters required to output each characteristic channel is reduced due to the separation of the three-dimensional convolution kernels. Specifically, d ² is used to represent the space size of the three-dimensional convolution kernel, n is used to represent the number of input channels of the three-dimensional convolution network, and the number of parameters is reduced from d ² ×n to d ² +n.

According to the embodiment of the disclosure, the initial browsing feature is enhanced, so that the dimension of the obtained browsing feature is increased, and the accuracy of the browsing feature is improved.

Fig. 4 schematically shows a data flow diagram of a product recommendation method according to an embodiment of the present disclosure.

As shown in fig. 4, the product recommendation method includes a browse record set 10, an initial browse feature 11, a browse feature 12, a product data map model 20, product features 21, and product recommendation results 30.

According to an embodiment of the present disclosure, an initial browsing feature 11 is derived based on browsing the record set 10. The initial browsing feature 11 is feature enhanced to obtain browsing feature 12. Based on the product data map model 20, product features 21 are obtained. Product recommendations 30 are obtained using the browse features 12 and the product features 21.

Fig. 5 schematically illustrates a schematic diagram of a three-dimensional convolution network according to a specific embodiment of this disclosure.

As shown in fig. 5, the three-dimensional convolution network includes a plurality of three-dimensional convolution layers. For example, the plurality of three-dimensional convolution layers includes three-dimensional convolution layer 510, three-dimensional convolution layers 520, …, three-dimensional convolution layer 5P0. Each three-dimensional convolution layer includes a one-dimensional convolution layer and a two-dimensional convolution layer. For example, the three-dimensional convolution layer 510 includes a one-dimensional convolution layer 510_1 and a two-dimensional convolution layer 510_2. The three-dimensional convolution layer 520 includes a one-dimensional convolution layer 520_1 and a two-dimensional convolution layer 520_2. To do so, the three-dimensional convolution layer 5P0 includes a one-dimensional convolution layer 5p0_1 and a two-dimensional convolution layer 5p0_2.

After the initial browsing feature 11 is input to the three-dimensional convolution layer 510, the intermediate browsing feature output by the three-dimensional convolution layer 510 is input to the three-dimensional convolution layer 320, and so on, until the three-dimensional convolution layer 5P0 outputs the browsing feature 12, in accordance with an embodiment of the present disclosure.

According to an embodiment of the present disclosure, the initial browsing feature includes a first sub-feature, a second sub-feature, and a third sub-feature.

According to an embodiment of the present disclosure, determining an initial browsing characteristic of a target user based on a browsing record set includes: determining a first data set from the browsing record set, and performing vectorization representation to obtain a first sub-feature, wherein the browsing amount of the products in the first data set in a first history period is higher than the browsing amount of other products in the browsing record in the first history period; determining a second data set from the browsing record set, and performing vectorization representation to obtain a second sub-feature, wherein the browsing amount of the products in the second product set in a second history period is higher than the browsing amount of other products in the browsing record in the second history period, and the second history period is longer than the first history period; determining the respective browsing times of each product category from the browsing record set, and performing vectorization representation to obtain a third sub-feature; and obtaining the initial browsing data characteristic based on the first sub-characteristic, the second sub-characteristic and the third sub-characteristic.

According to embodiments of the present disclosure, the products in the first dataset may be stored in the form of product identifications. When determining the first data set, product identifiers of a plurality of products with highest browsing amounts of the target user in the first historical period can be determined in the browsing record set, and the first data set is generated according to the product identifiers. Specifically, the first data set may be product identifications of six products that the target user browses the most in the last three days.

According to embodiments of the present disclosure, the products in the second data set may also be stored in the form of product identifications. When determining the second data set, product identifiers of a plurality of products with highest browsing amounts of the target user in the second history period can be determined in the browsing record set, and the second data set is generated according to the product identifiers. In particular, the second data set may be product identifications of the five products with the greatest amount of browsing for the target user since registering the account.

According to embodiments of the present disclosure, since the second history period is longer than the first history period, the products in the second data set may characterize the long-term preferred products of the target user, the products in the first data set may characterize the recent preferred products of the target user, and further the change in the preferences of the target user may be captured.

According to an embodiment of the present disclosure, the third sub-feature is associated with the target user's preference for the product category. Specifically, the browsing times of the target user on different product categories can be determined by analyzing the browsing record of the target user in one month, so that the third sub-feature is determined.

According to the embodiment of the disclosure, the data processing speed can be improved by vectorizing the first data set, the second data set and the browsing times.

According to the embodiment of the present disclosure, the initial browsing feature may be obtained by vector-stitching the first sub-feature, the second sub-feature, and the third sub-feature, but is not limited thereto. Different weights can be respectively configured for the first sub-feature, the second sub-feature and the third sub-feature according to service requirements, and the importance degree of the different sub-features can be adjusted.

According to the embodiment of the disclosure, the browsing characteristics of the target user are divided into different sub-characteristics, so that the browsing characteristics reflect the preference of the target user from different aspects, and further more accurate product recommendation results are obtained.

According to an embodiment of the present disclosure, determining respective browsing times of each product category from a browsing record set, and performing vectorized representation to obtain a third sub-feature, including: determining the browsing times of a target user aiming at each product category based on the browsing record set; vectorizing the multiple browsing times according to the preset arrangement sequence of each product category to generate an initial third sub-feature; and carrying out normalization operation on the initial third sub-feature to obtain a third sub-feature.

According to the embodiment of the disclosure, the product categories are all the product categories, and the browsing times are 0 for the product categories that the target user does not browse. Specifically, when the product is a financial product, the product comprises a regular deposit, a national debt, a monetary fund, insurance and stock, and the target user browses the regular deposit 3 times, browses the national debt 2 times, browses the monetary fund 0 times, browses the insurance 5 times and browses the stock 0 times in the last month.

According to the embodiment of the disclosure, the preset arrangement sequence may represent the arrangement sequence of the browsing times of different product categories in the initial third sub-feature. For example, the predetermined ranking order is regular deposit, national debt, monetary fund, insurance, stock, then the initial third sub-feature is (3,2,0,5,0).

According to the embodiment of the disclosure, in the case that the browsing records in the browsing record set are more, the number of browsing times in the initial third sub-feature is larger, thereby reducing the processing efficiency. And because the third sub-feature is used for representing the preference of the target user for the product category, the preference of the target user for the product category can be determined only by paying attention to the ratio of browsing times of different product categories, so that the processing efficiency can be improved by carrying out normalization processing on the initial third sub-feature, and meanwhile, the preference information of the target user for the product category is reserved. Specifically, the normalized initial third sub-feature is (0.3,0.2,0,0.5,0).

According to the embodiment of the disclosure, the data processing efficiency is improved by normalizing the initial third sub-feature.

According to an embodiment of the present disclosure, generating product features based on a product data graph model includes: generating an adjacency matrix of the product data graph model; obtaining a degree matrix of the product data graph model based on the adjacency matrix; and obtaining product features based on the adjacency matrix and the degree matrix.

According to embodiments of the present disclosure, the adjacency matrix may be determined by the side information of the product graph model. Since the side information characterizes that two products corresponding to two product nodes connected to the side have the same or similar product attributes, adjacent nodes in the adjacency matrix also have the same or similar product attributes.

According to embodiments of the present disclosure, a degree matrix may be derived from an adjacency matrix, which may characterize to some extent the degree of similarity between two products having the same or similar product attributes.

According to embodiments of the present disclosure, product features derived based on the adjacency matrix and the degree matrix may characterize the similarity between products. Specifically, the product features can be obtained by directly performing matrix splicing on the adjacent matrix and the degree matrix, or can be obtained by calculating the adjacent matrix and the degree matrix, and the method is not limited in any way.

According to the embodiment of the disclosure, the product characteristics are obtained through the adjacency matrix and the degree matrix, so that the cold door product can be covered, and meanwhile, the product characteristics are more accurate.

According to an embodiment of the present disclosure, a product recommendation network includes at least one graph roll layer and a map layer stacked.

According to an embodiment of the present disclosure, inputting browsing features and product features into a product recommendation network, generating product recommendation results for a target user, includes: inputting the browsing features and the product features into at least one graph convolution layer to obtain convolution features; inputting the convolution characteristics into a mapping layer, and determining the purchase probability of each product related to the target user purchasing product data graph model; and determining a product recommendation result based on the respective purchase probabilities of the plurality of products.

According to the embodiment of the disclosure, a product recommendation network may include one graph convolution layer or may include multiple graph convolution layers, which is not limited in any way. Specifically, two layers of the drawing volume may be included.

According to the embodiment of the disclosure, the browsing features and the product features are convolved through the graph convolution layer, so that the obtained convolution features are more accurate.

According to embodiments of the present disclosure, the mapping layer may include a normalized exponential function (softmax). And classifying the output characteristics of the picture scroll lamination by using a softmax function, and taking the matching degree of the output characteristics and each product as the purchase probability of purchasing the product by a target user.

According to the embodiment of the disclosure, when the mapping layer classifies the output characteristics, the number of categories can be set to be equal to the number of products, all the products are covered, and the recommended waterfall effect is avoided.

According to the embodiment of the disclosure, the product features and the browsing features are processed through graph convolution, so that the obtained product recommendation result is more accurate.

According to an embodiment of the present disclosure, each graph volume layer includes a convolution kernel and an activation function; inputting the browsing feature and the product feature into at least one graph roll layer to obtain a convolution feature, wherein the method comprises the following steps: inputting the ith convolution feature output by the ith picture convolution layer into the ith+1 convolution kernel aiming at the ith+1 picture convolution layer to obtain the ith+1 initial convolution feature, wherein i is greater than or equal to 1; and inputting the (i+1) th initial convolution feature into the (i+1) th activation function to obtain the (i+1) th convolution feature.

According to an embodiment of the present disclosure, the processing procedure of the graph convolution layer may be shown in the following formula (2):

Wherein A is an adjacency matrix, D is a degree matrix, H is a convolution feature, W is a convolution kernel, and sigma is an activation function.

Fig. 6 schematically illustrates a schematic diagram of stacked multiple graph convolutional layers, in accordance with an embodiment of the present disclosure.

As shown in fig. 6, the plurality of graph convolution layers includes graph convolution layer 610, graph convolution layers 620, …, and graph convolution layer 6N0. Each graph convolution layer includes a convolution kernel and an activation function. For example, the gallery stack 610 includes a convolution kernel 610_1 and an activation function 610_2. The graph convolution layer 620 includes a convolution kernel 620_1 and an activation function 620_2. To do so, the picture scroll layer 6N0 includes a convolution kernel 6n0_1 and an activation function 6n0_2.

After the browse feature 12 and the product feature 21 are input to the convolution layer 610, the intermediate convolution feature output by the convolution layer 610 is input to the convolution layer 620, and so on, until the convolution layer 6N0 outputs the convolution feature 40, in accordance with an embodiment of the present disclosure.

Fig. 7 schematically illustrates a flow chart of a product recommendation network training method according to an embodiment of the present disclosure.

As shown in fig. 7, the product recommendation network training method includes operations S710 to S740.

In operation S710, sample browsing features of the sample user are obtained based on the sample browsing record set of the sample user.

In operation S720, sample product features are generated based on the sample product data graph model.

In operation S730, the sample browsing characteristics and the sample product characteristics are input into the product recommendation network, and a sample product recommendation result for the sample user is generated.

In operation S740, the product recommendation network is trained based on the sample product recommendation results and the tags matched with the sample browsing record set.

According to embodiments of the present disclosure, the label includes the sample user's product results of interest. Specifically, the product of interest may be a product purchased by a sample user, or a product that is used by a sample to collect and join a shopping cart, which is not limited in any way.

According to embodiments of the present disclosure, for each sample product, a browsing record of a sample user who purchased the sample product may be obtained as a sample browsing record set.

It is reasonable in accordance with the present disclosure that the sample browsing feature may likewise include a first sample sub-feature, a second sample sub-feature, and a third sample sub-feature. Specifically, 100 sample users purchase the product 1, 10 products browsed before each sample user purchases the product 1 are obtained, 1000 products are obtained, 6 products with highest occurrence frequency are screened out of the 1000 products, and the first sample sub-feature is generated; obtaining products 5 before browsing times of the 100 sample users in the last month to obtain 500 products, and selecting 5 products with highest occurrence frequency from the 500 products to obtain second sample sub-features; and acquiring the browsing times of 100 sample users for each product category in 1 month, and determining a third sample sub-feature according to the browsing times.

According to the embodiment of the disclosure, the product recommendation network is trained through the interesting products of the sample user, so that the obtained product recommendation result is more accurate.

Based on the product recommendation method, the disclosure further provides a product recommendation device. The device will be described in detail below in connection with fig. 8.

Fig. 8 schematically illustrates a block diagram of a product recommendation device according to an embodiment of the present disclosure.

As shown in fig. 8, the product recommendation device 800 of this embodiment includes a browsing feature generation module 810, a product feature generation module 820, and a recommendation result generation module 830.

The browsing feature generation module 810 is configured to obtain browsing features of the target user based on the browsing record set of the target user. In an embodiment, the browsing feature generation module 810 may be configured to perform the operation S210 described above, which is not described herein.

The product feature generation module 820 is configured to generate product features based on the product data graph model. In an embodiment, the product feature generating module 820 may be used to perform the operation S220 described above, which is not described herein.

The recommendation result generation module 830 is configured to input the browsing feature and the product feature into a product recommendation network, and generate a product recommendation result for a target user. In an embodiment, the recommendation result generation module 830 may be configured to perform the operation S230 described above, which is not described herein.

According to the embodiment of the disclosure, the product recommendation network is obtained by training a graph convolution neural network by using a product data graph model, the product data graph model comprises a plurality of nodes and a plurality of edges, each node corresponds to a product, node information of each node comprises a product identifier, and each edge information of each edge represents an association relationship between two nodes corresponding to the edge information.

According to an embodiment of the present disclosure, the recommendation result generation module 830 includes a convolution sub-module, a probability determination sub-module, and a result determination sub-module.

And the convolution submodule is used for inputting the browsing characteristics and the product characteristics into at least one graph convolution layer to obtain the convolution characteristics.

And the probability determination submodule is used for inputting the convolution characteristics into the mapping layer and determining the purchase probability of each product related to the target user purchasing the product data graph model.

And the result determining submodule is used for determining a product recommendation result based on the purchase probability of each of the plurality of products.

According to an embodiment of the present disclosure, the browse feature generation module 810 includes an initial feature generation sub-module and a feature enhancer module.

And the initial characteristic generation sub-module is used for determining the initial browsing characteristics of the target user based on the browsing record set.

The feature enhancement sub-module is used for inputting the initial browsing features into a preset three-dimensional convolution network, carrying out feature enhancement on the initial browsing features and generating browsing features.

According to an embodiment of the present disclosure, an initial feature generation submodule includes a first generation unit, a second generation unit, a third generation unit, and an initial generation unit.

The first generation unit is used for determining a first data set from the browsing record set and carrying out vectorization representation to obtain a first sub-feature, wherein the browsing amount of the products in the first data set in the first history period is higher than the browsing amount of other products in the browsing record in the first history period.

The second generating unit is used for determining a second data set from the browsing record set and carrying out vectorization representation to obtain a second sub-feature, wherein the browsing amount of the products in the second product set in a second history period is higher than the browsing amount of other products in the browsing record in a second history period, and the second history period is longer than the first history period.

And the third generating unit is used for determining the respective browsing times of each product category from the browsing record set, and carrying out vectorization representation to obtain a third sub-feature.

The initial generation unit is used for obtaining initial browsing data characteristics based on the first sub-characteristics, the second sub-characteristics and the third sub-characteristics.

According to an embodiment of the present disclosure, the product feature generation module 820 includes a first matrix generation sub-module, a second matrix generation sub-module, and a product feature generation sub-module.

The first matrix generation module is used for generating an adjacency matrix of the product data graph model.

The second matrix generation module is used for obtaining a degree matrix of the product data graph model based on the adjacent matrix.

The product feature generation submodule is used for obtaining product features based on the adjacency matrix and the degree matrix.

According to an embodiment of the present disclosure, the third generating unit further includes a number determining subunit, an arranging subunit, and a normalizing subunit.

The number-of-times determining subunit is configured to determine, based on the browsing record set, the number of times of browsing by the target user for each product category.

The arrangement sub-unit is used for vectorizing the plurality of browsing times according to the preset arrangement sequence of each product category, and generating an initial third sub-feature.

And the normalization sub-unit is used for performing normalization operation on the initial third sub-feature to obtain the third sub-feature.

Any of the browsing feature generation module 810, the product feature generation module 820, and the recommendation result generation module 830 may be combined in one module to be implemented, or any of them may be split into multiple modules, according to embodiments of the present disclosure. Or at least some of the functionality of one or more of the modules may be combined with, and implemented in, at least some of the functionality of other modules. At least one of the browsing feature generation module 810, the product feature generation module 820, and the recommendation result generation module 830 may be implemented, at least in part, as hardware circuitry, such as a Field Programmable Gate Array (FPGA), a Programmable Logic Array (PLA), a system on a chip, a system on a substrate, a system on a package, an Application Specific Integrated Circuit (ASIC), or in hardware or firmware, such as any other reasonable way of integrating or packaging the circuitry, or in any one of or a suitable combination of any of the three. Or at least one of the browsing characteristic generation module 810, the product characteristic generation module 820 and the recommendation result generation module 830 may be at least partially implemented as a computer program module which, when executed, may perform the corresponding functions.

Based on the product recommendation network training method, the disclosure also provides a product recommendation network training device. The device will be described in detail below in connection with fig. 9.

Fig. 9 schematically illustrates a block diagram of a product recommendation network training device according to an embodiment of the present disclosure.

As shown in fig. 9, the product recommendation network training apparatus 900 of this embodiment includes a sample browsing feature generation module 910, a sample product feature generation module 920, a sample recommendation result generation module 930, and a model training module 940.

The sample browsing feature generation module 910 is configured to obtain sample browsing features of a sample user based on a sample browsing record set of the sample user. In an embodiment, the sample browsing feature generation module 910 may be configured to perform the operation S710 described above, which is not described herein.

The sample product feature generation module 920 is configured to generate sample product features based on the sample product data graph model. In an embodiment, the sample product feature generating module 920 may be configured to perform the operation S720 described above, which is not described herein.

The sample recommendation result generating module 930 is configured to input the sample browsing feature and the sample product feature into a product recommendation network, and generate a sample product recommendation result for the sample user, where the product data graph model includes a plurality of product nodes and a plurality of edges, node information of each of the plurality of product nodes includes product attribute information, and edge information of each of the plurality of edges characterizes that two products corresponding to two product nodes connected to the edge have a product association relationship. In an embodiment, the sample recommendation result generation module 930 may be configured to perform the operation S730 described above, which is not described herein.

The model training module 940 is configured to train the product recommendation network based on the sample product recommendation results and the labels that match the sample browsing record set, where the labels include the product results of interest of the sample user. In an embodiment, the model training module 940 may be configured to perform the operation S740 described above, which is not described herein.

As shown in fig. 10, an electronic device 1000 according to an embodiment of the present disclosure includes a processor 1001 that can perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM) 1002 or a program loaded from a storage section 1008 into a Random Access Memory (RAM) 1003. The processor 1001 may include, for example, a general purpose microprocessor (e.g., a CPU), an instruction set processor and/or an associated chipset and/or a special purpose microprocessor (e.g., an Application Specific Integrated Circuit (ASIC)), or the like. The processor 1001 may also include on-board memory for caching purposes. The processor 1001 may include a single processing unit or multiple processing units for performing different actions of the method flows according to embodiments of the present disclosure.

In the RAM 1003, various programs and data necessary for the operation of the electronic apparatus 1000 are stored. The processor 1001, the ROM 1002, and the RAM 1003 are connected to each other by a bus 1004. The processor 1001 performs various operations of the method flow according to the embodiment of the present disclosure by executing programs in the ROM 1002 and/or the RAM 1003. Note that the program may be stored in one or more memories other than the ROM 1002 and the RAM 1003. The processor 1001 may also perform various operations of the method flow according to the embodiments of the present disclosure by executing programs stored in the one or more memories.

According to an embodiment of the disclosure, the electronic device 1000 may also include an input/output (I/O) interface 1005, the input/output (I/O) interface 1005 also being connected to the bus 1004. The electronic device 1000 may also include one or more of the following components connected to the I/O interface 1005: an input section 1006 including a keyboard, a mouse, and the like; an output portion 1007 including a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), etc., and a speaker, etc.; a storage portion 1008 including a hard disk or the like; and a communication section 1009 including a network interface card such as a LAN card, a modem, or the like. The communication section 1009 performs communication processing via a network such as the internet. The drive 1010 is also connected to the I/O interface 1005 as needed. A removable medium 1011, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like, is installed as needed in the drive 1010, so that a computer program read out therefrom is installed as needed in the storage section 1008.

The present disclosure also provides a computer-readable storage medium that may be embodied in the apparatus/device/system described in the above embodiments; or may exist alone without being assembled into the apparatus/device/system. The computer-readable storage medium carries one or more programs which, when executed, implement methods in accordance with embodiments of the present disclosure.

According to embodiments of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium, which may include, for example, but is not limited to: a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this disclosure, a computer-readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. For example, according to embodiments of the present disclosure, the computer-readable storage medium may include ROM 1002 and/or RAM 1003 and/or one or more memories other than ROM 1002 and RAM 1003 described above.

Embodiments of the present disclosure also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowcharts. The program code, when executed in a computer system, causes the computer system to implement the item recommendation method provided by embodiments of the present disclosure.

The above-described functions defined in the system/apparatus of the embodiments of the present disclosure are performed when the computer program is executed by the processor 1001. The systems, apparatus, modules, units, etc. described above may be implemented by computer program modules according to embodiments of the disclosure.

In one embodiment, the computer program may be based on a tangible storage medium such as an optical storage device, a magnetic storage device, or the like. In another embodiment, the computer program may also be transmitted in the form of signals on a network medium, distributed, and downloaded and installed via the communication section 1009, and/or installed from the removable medium 1011. The computer program may include program code that may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the foregoing.

In such an embodiment, the computer program may be downloaded and installed from a network via the communication portion 1009, and/or installed from the removable medium 1011. The above-described functions defined in the system of the embodiments of the present disclosure are performed when the computer program is executed by the processor 1001. The systems, devices, apparatus, modules, units, etc. described above may be implemented by computer program modules according to embodiments of the disclosure.

According to embodiments of the present disclosure, program code for performing computer programs provided by embodiments of the present disclosure may be written in any combination of one or more programming languages, and in particular, such computer programs may be implemented in high-level procedural and/or object-oriented programming languages, and/or assembly/machine languages. Programming languages include, but are not limited to, such as Java, c++, python, "C" or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, partly on a remote computing device, or entirely on the remote computing device or server. In the case of remote computing devices, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., connected via the Internet using an Internet service provider).

The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams or flowchart illustration, and combinations of blocks in the block diagrams or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

Those skilled in the art will appreciate that the features recited in the various embodiments of the disclosure and/or in the claims may be provided in a variety of combinations and/or combinations, even if such combinations or combinations are not explicitly recited in the disclosure. In particular, the features recited in the various embodiments of the present disclosure and/or the claims may be variously combined and/or combined without departing from the spirit and teachings of the present disclosure. All such combinations and/or combinations fall within the scope of the present disclosure.

The embodiments of the present disclosure are described above. These examples are for illustrative purposes only and are not intended to limit the scope of the present disclosure. Although the embodiments are described above separately, this does not mean that the measures in the embodiments cannot be used advantageously in combination. The scope of the disclosure is defined by the appended claims and equivalents thereof. Various alternatives and modifications can be made by those skilled in the art without departing from the scope of the disclosure, and such alternatives and modifications are intended to fall within the scope of the disclosure.

Claims

1. A method of product recommendation, the method comprising:

obtaining browsing characteristics of a target user based on a browsing record set of the target user;

generating product features based on the product data graph model; and

Inputting the browsing characteristics and the product characteristics into a product recommendation network, generating a product recommendation result for the target user,

The product data graph model comprises a plurality of product nodes and a plurality of edges, the node information of each product node comprises product attribute information, and the edge information of each edge represents a product association relationship between two products corresponding to two product nodes connected with the edge.

2. The method of claim 1, wherein the product recommendation network comprises at least one of a stacked gallery and map layer,

The step of inputting the browsing characteristics and the product characteristics into a product recommendation network to generate a product recommendation result for the target user comprises the following steps:

Inputting the browsing features and the product features to the at least one picture scroll layer to obtain convolution features;

inputting the convolution characteristics into the mapping layer, and determining the purchase probability of each product related to the target user purchasing the product data graph model; and

And determining the product recommendation result based on the purchase probabilities of the products.

3. The method of claim 1, wherein generating browsing characteristics of the target user based on the target user's browsing record set comprises:

determining initial browsing characteristics of the target user based on the browsing record set; and

Inputting the initial browsing characteristics into a preset three-dimensional convolution network, carrying out characteristic enhancement on the initial browsing characteristics, and generating the browsing characteristics.

4. The method of claim 3, wherein the initial browsing feature comprises a first sub-feature, a second sub-feature and a third sub-feature,

The determining, based on the browsing record set, an initial browsing feature of the target user includes:

Determining a first data set from the browsing record set, and performing vectorization representation to obtain the first sub-feature, wherein the browsing amount of the products in the first data set in a first history period is higher than the browsing amount of other products in the browsing record in the first history period;

Determining a second data set from the browsing record set, and performing vectorization representation to obtain the second sub-feature, wherein the browsing amount of the products in the second product set in a second history period is higher than the browsing amount of other products in the browsing record in the second history period, and the second history period is longer than the first history period;

Determining the respective browsing times of each product category from the browsing record set, and performing vectorization representation to obtain the third sub-feature; and

And obtaining the initial browsing data characteristic based on the first sub-characteristic, the second sub-characteristic and the third sub-characteristic.

5. The method of claim 1, wherein generating product features based on the product data graph model comprises:

Generating an adjacency matrix of the product data graph model;

obtaining a degree matrix of the product data graph model based on the adjacency matrix; and

And obtaining the product characteristics based on the adjacency matrix and the degree matrix.

6. The method of claim 5, wherein determining the respective number of views for each product category from the set of view records and performing a vectorized representation to obtain the third sub-feature comprises:

Determining the browsing times of the target user for each product category based on the browsing record set;

Vectorizing the browsing times according to the preset arrangement sequence of each product category to generate an initial third sub-feature;

And normalizing the initial third sub-feature to obtain the third sub-feature.

7. A method for product recommendation network training, the method comprising:

based on a sample browsing record set of a sample user, obtaining sample browsing characteristics of the sample user;

generating sample product features based on the sample product data graph model;

Inputting the sample browsing characteristics and the sample product characteristics into a product recommendation network, and generating a sample product recommendation result for the sample user, wherein the product data graph model comprises a plurality of product nodes and a plurality of edges, node information of each of the plurality of product nodes comprises product attribute information, and edge information of each of the plurality of edges represents a product association relationship between two products corresponding to two product nodes connected with the edge; and

Training the product recommendation network based on the sample product recommendation results and a tag matched with the sample browsing record set, wherein the tag comprises the product results of interest of the sample user.

8. A product recommendation device, the device comprising:

the browsing characteristic generation module is used for obtaining the browsing characteristics of the target user based on the browsing record set of the target user;

the product characteristic generating module is used for generating product characteristics based on the product data graph model; and

A recommendation result generation module for inputting the browsing characteristics and the product characteristics into a product recommendation network to generate a product recommendation result for the target user,

The product recommendation network is obtained by training a graph convolution neural network by using a product data graph model, the product data graph model comprises a plurality of nodes and a plurality of edges, each node corresponds to a product, node information of each node comprises a product identifier, and each edge information of each edge represents an association relationship between two nodes corresponding to the edge information.

9. A product recommendation network training device, the device comprising:

The sample browsing feature generation module is used for obtaining sample browsing features of the sample user based on a sample browsing record set of the sample user;

the sample product feature generation module is used for generating sample product features based on the sample product data graph model;

The sample recommendation result generation module is used for inputting the sample browsing characteristics and the sample product characteristics into a product recommendation network to generate sample product recommendation results for the sample users, wherein the product data graph model comprises a plurality of product nodes and a plurality of edges, node information of each of the plurality of product nodes comprises product attribute information, and edge information of each of the plurality of edges represents a product association relationship between two products corresponding to two product nodes connected with the edges; and

And the model training module is used for training the product recommendation network based on the sample product recommendation result and a label matched with the sample browsing record set, wherein the label comprises the interesting product result of the sample user.

10. An electronic device, comprising:

One or more processors;

a memory for storing one or more computer programs,

Characterized in that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 7.

11. A computer-readable storage medium, on which a computer program or instructions is stored, characterized in that the computer program or instructions, when executed by a processor, implement the steps of the method according to any one of claims 1-7.

12. A computer program product comprising a computer program or instructions which, when executed by a processor, implement the steps of the method according to any one of claims 1 to 7.