Commodity recommendation method and device, nonvolatile storage medium and computer equipment

By constructing interactive graphs and using neural network models, we can deeply explore the product characteristics that are of interest to users, and solve the problem of low recommendation efficiency in the existing technology, and achieve accurate and efficient product recommendations.

CN119006096APending Publication Date: 2024-11-22CHINA TELECOM BESTPAY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411025033.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-29
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

Existing product recommendation methods cannot deeply explore the product characteristics that users are interested in, resulting in low recommendation efficiency.

Method used

By obtaining the characteristics of the target object and product, building an interactive graph and sorting the data using a pre-trained neural network model, products that meet user interests are recommended.

Benefits of technology

It realizes accurate and efficient recommendations of products that are of interest to users and improves recommendation efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119006096A_ABST
    Figure CN119006096A_ABST
Patent Text Reader

Abstract

The invention discloses a commodity recommendation method and device, a nonvolatile storage medium and computer equipment. The method comprises the steps that object features of a target object and respective commodity features of a plurality of commodities are acquired, and the commodity features comprise features of shops to which the commodities belong; combining the object features with respective commodity features of a plurality of commodities to obtain a plurality of groups of input data; the input data are input into a pre-trained neural network model, the input data are sorted, each training sample comprises features of a sample object and features of a sample commodity, and the multiple training samples comprise positive samples with interaction behaviors; features of sample objects included in the positive samples are determined by interaction behaviors among the sample objects, sample commodities and sample shops; and recommending commodities to the target object according to the sorted multiple groups of input data. The technical problem that the commodity recommendation efficiency is not high due to the fact that a traditional method cannot deeply perform feature mining on commodities interested by users is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of commodity recommendation, and in particular to a commodity recommendation method, device, non-volatile storage medium and computer equipment. Background Art

[0002] The personalized recommendation format has played an important role in local life scenarios. Different users have different preferences for different products and different stores. For example, some users pay more attention to the quality of the product and will choose high-quality products for consumption; some users pay more attention to the price of the product and tend to choose cost-effective products for consumption. In addition, users have different consumption habits in different time periods. For example, in the afternoon, some users prefer to enjoy afternoon tea with friends, while some users prefer leisure and entertainment activities. However, the existing product recommendation methods simply construct corresponding features for users, stores, products, and time periods for learning, and cannot deeply mine the features of products that users are interested in, resulting in low efficiency in product recommendation.

[0003] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0004] The embodiments of the present invention provide a product recommendation method, apparatus, non-volatile storage medium and computer device to at least solve the technical problem that the traditional method cannot deeply mine the features of the products that the user is interested in, resulting in low product recommendation efficiency.

[0005] According to one aspect of an embodiment of the present invention, a product recommendation method is provided, comprising: obtaining object features of a target object and product features of multiple products, wherein the product features of the multiple products include features of the store to which the products belong; combining the object features with the product features of the multiple products to obtain multiple groups of input data; inputting the multiple groups of input data into a pre-trained neural network model, and sorting the multiple groups of input data by the neural network model, wherein the neural network model is obtained by training an original neural network model with multiple training samples, each of the multiple training samples includes features of a sample object and features of a sample product, and the multiple training samples include positive samples with interactive behaviors, and features of the sample objects included in the positive samples are determined by the interactive behaviors among the sample object, the sample product and the sample store; and recommending products to the target object based on the sorted multiple groups of input data.

[0006] Optionally, the neural network model is trained in the following manner: obtaining initial features of the sample object and initial features of the sample product; determining a target sample product that interacts with the sample object; updating the initial features of the sample object based on the interaction between the sample object and the target sample product to obtain features of the sample object, and updating the initial features of the target sample product to obtain features of the target sample product; generating positive samples based on the features of the sample object and the features of the target sample product; and training the original neural network model based on the positive samples to obtain a neural network model.

[0007] Optionally, according to the interaction behavior between the sample object and the target sample product, the initial features of the sample object are updated to obtain the features of the sample object, and the initial features of the target sample product are updated to obtain the features of the target sample product, including: obtaining the initial features of the target sample store that interacts with the sample object; constructing a product interaction graph and a store interaction graph according to the interaction behavior between the sample object and the target sample product, and the interaction behavior between the sample object and the target sample store, wherein the product interaction graph includes sample object nodes, target sample product nodes, target sample store nodes, edges between the sample object nodes and the target sample product nodes, and edges between the target sample product nodes and the target sample store nodes, and the store interaction graph includes sample object nodes, target sample product nodes, target sample store nodes, edges between the sample object nodes and the target sample store nodes, and edges between the target sample product nodes and the target sample store nodes; based on the product interaction graph, the sample object nodes are updated. The initial characteristics of the sample object of the point are updated to obtain the first characteristics of the sample object, and the initial characteristics of the target sample commodity of the target sample commodity node are updated to obtain the first characteristics of the target sample commodity, and the initial characteristics of the target sample store of the target sample store node are updated to obtain the first characteristics of the target sample store; based on the store interaction graph, the initial characteristics of the sample object of the sample object node are updated to obtain the second characteristics of the sample object, and the initial characteristics of the target sample commodity of the target sample commodity node are updated to obtain the second characteristics of the target sample commodity, and the initial characteristics of the target sample store of the target sample store node are updated to obtain the second characteristics of the target sample store; according to the first characteristics of the sample object and the second characteristics of the sample object, the characteristics of the sample object are obtained, and according to the first characteristics of the target sample commodity, the second characteristics of the target sample commodity, the first characteristics of the target sample store and the second characteristics of the target sample store, the characteristics of the target sample commodity are obtained.

[0008] Optionally, based on the commodity interaction graph, the initial features of the sample object of the sample object node are updated to obtain the first feature of the sample object, and the initial features of the target sample commodity of the target sample commodity node are updated to obtain the first feature of the target sample commodity, and the initial features of the target sample store of the target sample store node are updated to obtain the first feature of the target sample store, including: based on the commodity interaction graph, the initial features of the sample object are updated according to the features of the nodes adjacent to the sample object node, and the initial features of the target sample commodity are updated according to the features of the nodes adjacent to the target sample commodity node, and the initial features of the target sample store are updated according to the features of the nodes adjacent to the target sample store node to obtain the commodity interaction graph of the updated features; attention extraction is performed on the commodity interaction graph of the updated features to determine the weights of the edges in the commodity interaction graph of the updated features; according to the weights of the edges in the commodity interaction graph of the updated features, the features of the nodes in the commodity interaction graph of the updated features are updated again to obtain the first feature of the sample object, the first feature of the target sample commodity, and the first feature of the target sample store.

[0009] Optionally, based on the store interaction graph, the initial features of the sample object of the sample object node are updated to obtain the second features of the sample object, and the initial features of the target sample commodity of the target sample commodity node are updated to obtain the second features of the target sample commodity, and the initial features of the target sample shop of the target sample shop node are updated to obtain the second features of the target sample shop, including: based on the store interaction graph, the initial features of the sample object are updated according to the features of the nodes adjacent to the sample object node, and the initial features of the target sample commodity are updated according to the features of the nodes adjacent to the target sample commodity node, and the initial features of the target sample shop are updated according to the features of the nodes adjacent to the target sample shop node to obtain the store interaction graph of the updated features; attention extraction is performed on the store interaction graph of the updated features to determine the weights of the edges in the store interaction graph of the updated features; according to the weights of the edges in the store interaction graph of the updated features, the features of the nodes in the store interaction graph of the updated features are updated again to obtain the second features of the sample object, the second features of the target sample commodity, and the second features of the target sample shop.

[0010] Optionally, based on the positive samples, the original neural network model is trained to obtain the neural network model, including: determining other sample products that have no interactive behavior with the sample object; generating negative samples based on the initial features of the other sample products and the features of the sample object; based on the positive samples and the negative samples, the original neural network model is trained to obtain the neural network model.

[0011] According to another aspect of an embodiment of the present invention, there is also provided a product recommendation device, comprising: an acquisition module, used to acquire object features of a target object, product features of each of a plurality of products, wherein the product features of each of the plurality of products include features of the store to which the products belong; a combination module, used to combine the object features with the product features of each of the plurality of products, respectively, to obtain multiple groups of input data; a training module, used to input the multiple groups of input data into a pre-trained neural network model, and the neural network model sorts the multiple groups of input data, wherein the neural network model is obtained by training an original neural network model with multiple training samples, each of the multiple training samples includes features of a sample object and features of a sample product, the multiple training samples include positive samples with interactive behaviors, and the features of the sample object included in the positive samples are determined by the interactive behaviors among the sample object, the sample product and the sample store; a recommendation module, used to recommend products to the target object based on the sorted multiple groups of input data.

[0012] According to another aspect of an embodiment of the present invention, a non-volatile storage medium is provided, the non-volatile storage medium including a stored program, wherein when the program is running, the device where the non-volatile storage medium is located is controlled to execute any one of the above-mentioned product recommendation methods.

[0013] According to yet another aspect of an embodiment of the present invention, a computer device is provided. The computer device includes a processor, and the processor is used to run a program. When the program is run, any one of the above-mentioned product recommendation methods is executed.

[0014] According to yet another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, any one of the above-mentioned product recommendation methods is implemented.

[0015] In an embodiment of the present invention, by acquiring object features of a target object and product features of multiple products, wherein the product features of multiple products include features of the store to which the products belong; combining the object features with the product features of multiple products respectively to obtain multiple groups of input data; inputting the multiple groups of input data into a pre-trained neural network model, and the neural network model sorts the multiple groups of input data, wherein the neural network model is obtained by training an original neural network model through multiple training samples, each of the multiple training samples includes features of a sample object and features of a sample product, and the multiple training samples include positive samples with interactive behaviors, and features of the sample object included in the positive samples are determined by the interactive behaviors among the sample object, the sample product and the sample store; recommending products to the target object based on the sorted multiple groups of input data, thereby achieving the purpose of deeply mining the features of the products that the user is interested in, thereby achieving the technical effect of accurately and efficiently recommending the products that the user is interested in, and further solving the technical problem that the traditional method cannot deeply mine the features of the products that the user is interested in, resulting in low efficiency in product recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0017] Figure 1 A hardware structure block diagram of a computer terminal for implementing a commodity recommendation method is shown;

[0018] Figure 2 is a flowchart of a commodity recommendation method provided according to an embodiment of the present invention;

[0019] Figure 3 is a schematic diagram of an object feature and a commodity feature extraction process provided according to an optional embodiment of the present invention;

[0020] Figure 4 is a schematic diagram of a commodity recommendation device provided according to an optional embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0022] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0023] First, some nouns or terms that appear in the description of the embodiments of the present application are subject to the following interpretations:

[0024] Attention extraction is a mechanism that enables neural network models to focus on important parts of input data.

[0025] According to an embodiment of the present invention, a method embodiment for product recommendation is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0026] The method embodiment provided in the first embodiment of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 FIG. 1 shows a hardware structure block diagram of a computer terminal for implementing a commodity recommendation method. Figure 1 As shown, the computer terminal 10 may include one or more (102a, 102b, ..., 102n are used to illustrate) processors (the processor may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0027] It should be noted that the one or more processors and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computer terminal 10. As described in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0028] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the commodity recommendation method in the embodiment of the present invention. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, realizing the commodity recommendation method of the above-mentioned application program. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0029] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 .

[0030] Figure 2 FIG. 1 is a flow chart of a commodity recommendation method according to an embodiment of the present invention. Figure 2 As shown, the method comprises the following steps:

[0031] Step S202, obtaining object features of the target object and product features of each of the plurality of products, wherein the product features of each of the plurality of products include features of the store to which the products belong.

[0032] In this step, the target object is the target user for this product recommendation. The object characteristics are some attributes related to the target user or his / her behavior, which can reflect the quality and characteristics of the target user from one or more aspects. They can be the target user's gender, age, height, weight, etc. Product characteristics are some attributes related to each type of product itself, which can reflect the quality and characteristics of each type of product from one or more aspects. They can be the price, category, purpose, raw materials, etc. Store characteristics are some attributes related to the store itself, which can reflect the quality and characteristics of the store from one or more aspects. They can be the store address, business hours, etc.

[0033] Step S204 , combining the object features with the commodity features of the plurality of commodities respectively, to obtain a plurality of groups of input data.

[0034] In this step, the object features are combined with the product features of each of the multiple products, and a set of input data includes the object features and the product features of one product. The combination can be to combine the separated object features and product features by splicing or other forms to form input data in vector form.

[0035] Step S206, inputting the multiple groups of input data into a pre-trained neural network model, and the neural network model sorts the multiple groups of input data, wherein the neural network model is obtained by training the original neural network model through multiple training samples, each of the multiple training samples includes the characteristics of the sample object and the characteristics of the sample product, and the multiple training samples include positive samples with interactive behaviors, and the characteristics of the sample objects included in the positive samples are determined by the interactive behaviors between the sample objects, the sample products and the sample stores.

[0036] In this step, pre-training refers to training the neural network model with multiple training samples before making product recommendations to find the optimal parameters of the neural network model. The neural network model is used to sort multiple sets of input data. The basis for sorting is the degree of interest of the target object in the product. It can be a recurrent neural network, Transformer, etc. Training samples refer to multiple sets of data used to train the neural network model, which are obtained by combining the characteristics of the sample object and the characteristics of the sample product. Positive samples refer to training samples in which there are interactive behaviors between sample objects, sample products and sample stores. Positive samples must be training samples, and training samples are not necessarily positive samples. Interactive behavior refers to the click, mark and other relationships between sample objects, sample products and sample stores. Sample objects, sample products and sample stores can be used as nodes in the graph, and the edges between the nodes can be used to characterize the interactive behaviors between sample objects, sample products and sample stores.

[0037] Step S208: recommending products to the target object based on the sorted multiple groups of input data.

[0038] In this step, the neural network model can be used to sort the multiple sets of output data, and the basis for sorting is the degree of interest of the target object in the product. Recommending products can be pushing one or more products that the target object is most interested in to the target object, and the push method can be to provide product information to the target object in the form of text messages, phone calls or emails.

[0039] Through the above steps, the purpose of deeply mining the features of the products that users are interested in is achieved, thereby achieving the technical effect of accurately and efficiently recommending the products of interest to users, thereby solving the technical problem that traditional methods cannot deeply mine the features of the products that users are interested in, resulting in low efficiency in product recommendations.

[0040] As an optional embodiment, the neural network model is trained in the following manner: obtaining the initial features of the sample object and the initial features of the sample product; determining the target sample product that has interactive behavior with the sample object; updating the initial features of the sample object according to the interactive behavior between the sample object and the target sample product to obtain the features of the sample object, and updating the initial features of the target sample product to obtain the features of the target sample product; generating positive samples according to the features of the sample object and the features of the target sample product; and training the original neural network model based on the positive samples to obtain the neural network model.

[0041] Optionally, obtaining the initial features of the sample object refers to obtaining some information that can characterize the target user, such as the user's basic information, wherein the information is fully authorized by the user in accordance with the regulations of the relevant region. Obtaining the initial features of the sample goods refers to obtaining some information that can characterize various types of goods, such as the store to which the goods belong, the name of the goods, the category, the price, the sales volume, etc. The interactive behavior can specifically be the buying, selling, clicking or marking behavior between the sample object and the sample goods. The initial features of the sample object and the target sample goods can be updated according to the interactive behavior between the sample object and the target sample goods. Since the features of the sample object and the features of the target sample goods are separate and independent, the features of the sample object and the features of the target sample goods can be spliced ​​to form a positive sample of the combination of the two. The positive sample can be used to train the neural network model so that the neural network model learns the knowledge to analyze the degree of interest of the target object in the target object goods.

[0042] As an optional embodiment, according to the interaction behavior between the sample object and the target sample product, the initial features of the sample object are updated to obtain the features of the sample object, and the initial features of the target sample product are updated to obtain the features of the target sample product, including: obtaining the initial features of the target sample store that interacts with the sample object; constructing a product interaction graph and a store interaction graph according to the interaction behavior between the sample object and the target sample product, and the interaction behavior between the sample object and the target sample store, wherein the product interaction graph includes sample object nodes, target sample product nodes, target sample store nodes, edges between the sample object nodes and the target sample product nodes, and edges between the target sample product nodes and the target sample store nodes, and the store interaction graph includes sample object nodes, target sample product nodes, target sample store nodes, edges between the sample object nodes and the target sample store nodes, and edges between the target sample product nodes and the target sample store nodes; based on the product interaction graph, the sample object is updated to obtain the features of the target sample product, and the initial features of the target sample store are updated to obtain the features of the target sample store. The initial characteristics of the sample object of the object node are updated to obtain the first characteristics of the sample object, and the initial characteristics of the target sample commodity of the target sample commodity node are updated to obtain the first characteristics of the target sample commodity, and the initial characteristics of the target sample store of the target sample store node are updated to obtain the first characteristics of the target sample store; based on the store interaction graph, the initial characteristics of the sample object of the sample object node are updated to obtain the second characteristics of the sample object, and the initial characteristics of the target sample commodity of the target sample commodity node are updated to obtain the second characteristics of the target sample commodity, and the initial characteristics of the target sample store of the target sample store node are updated to obtain the second characteristics of the target sample store; according to the first characteristics of the sample object and the second characteristics of the sample object, the characteristics of the sample object are obtained, and according to the first characteristics of the target sample commodity, the second characteristics of the target sample commodity, the first characteristics of the target sample store and the second characteristics of the target sample store, the characteristics of the target sample commodity are obtained.

[0043] Optionally, the initial features of the target sample product may include the features of the target sample store. The product interaction graph is composed of nodes and edges. The nodes are used to characterize the features, and the edges are used to characterize the interactive relationships between nodes, including sample object nodes, target sample product nodes, target sample store nodes, edges between sample object nodes and target sample product nodes, and edges between target sample product nodes and target sample store nodes. The store interaction graph is composed of nodes and edges. The nodes are used to characterize the features, and the edges are used to characterize the interactive relationships between nodes, including sample object nodes, target sample product nodes, target sample store nodes, edges between sample object nodes and target sample store nodes, and edges between target sample product nodes and target sample store nodes.

[0044] The difference between the product interaction graph and the store interaction graph is that the product interaction graph has edges between the sample object nodes and the target sample product nodes, while the store interaction graph does not. The store interaction graph has edges between the sample object nodes and the target sample store nodes, while the product interaction graph does not. The initial features of the sample object, the target sample product, and the target sample store are all independent of the interaction relationship. Based on the connection relationship between each node and each edge in the product interaction graph, the initial features of the sample object, the target sample product, and the target sample store are updated to obtain the first features of the sample object, the target sample product, and the target sample store that can characterize the interaction relationship; based on the connection relationship between each node and each edge in the store interaction graph, the initial features of the sample object, the target sample product, and the target sample store are updated to obtain the second features of the sample object, the target sample product, and the target sample store that can characterize the interaction relationship. The difference between the first feature and the second feature is that the first feature is obtained based on the interaction behavior in the product interaction graph, and the second feature is obtained based on the interaction behavior in the store interaction graph. The first feature of the sample object and the second feature of the sample object are concatenated to obtain the feature of the sample object, and the first feature of the target sample product, the second feature of the target sample product, the first feature of the target sample store, and the second feature of the target sample store are concatenated to obtain the feature of the target sample product.

[0045] Specifically, a product interaction graph G is constructed based on sample objects, target sample products, target sample stores, the interaction relationship between sample objects and target sample products, and the interaction relationship between target sample products and target sample stores. I , including sample object nodes Target sample product node Target sample store node The edge between the sample object node and the target sample product node The edge between the target sample product node and the target sample store node The product interaction graph G is constructed based on the sample objects, target sample products, target sample stores, the interaction relationship between the sample objects and the target sample stores, and the interaction relationship between the target sample products and the target sample stores. S , including sample object nodes Target sample product node Target sample store node The edge between the sample object node and the target sample store node The edge between the target sample product node and the target sample store node

[0046] It should be noted that the product interaction graph and store interaction graph can be further decomposed according to time periods. n When Includes sample object nodes Target sample product node Target sample store node The edge between the sample object node and the target sample product node The edge between the target sample product node and the target sample store node In time period T n When Includes sample object nodes Target sample product node Target sample store node The edge between the sample object node and the target sample store node The edge between the target sample product node and the target sample store node

[0047] As an optional embodiment, based on the commodity interaction graph, the initial features of the sample object of the sample object node are updated to obtain the first feature of the sample object, and the initial features of the target sample commodity of the target sample commodity node are updated to obtain the first feature of the target sample commodity, and the initial features of the target sample store of the target sample store node are updated to obtain the first feature of the target sample store, including: based on the commodity interaction graph, the initial features of the sample object are updated according to the features of the nodes adjacent to the sample object node, and the initial features of the target sample commodity are updated according to the features of the nodes adjacent to the target sample commodity node, and the initial features of the target sample store are updated according to the features of the nodes adjacent to the target sample store node to obtain the commodity interaction graph of the updated features; attention extraction is performed on the commodity interaction graph of the updated features to determine the weights of the edges in the commodity interaction graph of the updated features; according to the weights of the edges in the commodity interaction graph of the updated features, the features of the nodes in the commodity interaction graph of the updated features are updated again to obtain the first feature of the sample object, the first feature of the target sample commodity, and the first feature of the target sample store.

[0048] Optionally, the initial features of the sample object node, the target sample store node, and the target sample store node are irrelevant to the interactive behavior. Therefore, based on the connection relationship of the nodes and edges of the commodity interaction graph and the features of the adjacent nodes, the initial features of the sample object node, the target sample store node, and the target sample store node are updated so that the updated initial features are related to the interactive behavior. Since the sample object has different degrees of interest in different target commodity sample nodes, the weight of the corresponding edge in the commodity interaction graph is calculated based on the updated features of the target sample store node, and then the first feature of the sample object is calculated based on the weight of the edge in the obtained commodity interaction graph.

[0049] Specifically, based on the product interaction graph G I , according to the characteristics of the nodes adjacent to the sample object node, the initial characteristics of the sample object Update, and, according to the characteristics of the nodes adjacent to the target sample product node, update the initial characteristics of the target sample product Update, and, according to the characteristics of the nodes adjacent to the target sample store node, update the initial characteristics of the target sample store Update, expressed as

[0050]

[0051]

[0052] Among them, e is the index of the edge, |{*}| is the modulo symbol, which is used to calculate the number of variables in the array {*}. is the initial feature of the updated sample object, V s I′ is the initial feature of the updated target sample store, is the initial feature of the updated target sample product. Based on the updated features of the target sample store node, the first feature of the sample object is calculated, which is expressed as:

[0053]

[0054] in, is the first feature of the sample object.

[0055] It should be noted that in the time period T n When, based on the product interaction graph According to the features of the nodes adjacent to the sample object node, the initial features of the sample object are Update, and, according to the characteristics of the nodes adjacent to the target sample product node, update the initial characteristics of the target sample product Update, and, according to the characteristics of the nodes adjacent to the target sample store node, update the initial characteristics of the target sample store Update, expressed as

[0056]

[0057] in, is the updated sample object in time period T n The initial characteristics of is the updated target sample store in time period T n The initial characteristics of is the updated target sample product in time period T nBased on the updated features of the target sample store node, calculate the sample object in time period T n The first feature of is expressed as:

[0058]

[0059] in, is the sample object in time period T n The first feature.

[0060] As an optional embodiment, based on the store interaction graph, the initial features of the sample object of the sample object node are updated to obtain the second features of the sample object, and the initial features of the target sample commodity of the target sample commodity node are updated to obtain the second features of the target sample commodity, and the initial features of the target sample store of the target sample store node are updated to obtain the second features of the target sample store, including: based on the store interaction graph, the initial features of the sample object are updated according to the features of the nodes adjacent to the sample object node, and the initial features of the target sample commodity are updated according to the features of the nodes adjacent to the target sample commodity node, and the initial features of the target sample store are updated according to the features of the nodes adjacent to the target sample store node to obtain the store interaction graph of the updated features; attention extraction is performed on the store interaction graph of the updated features to determine the weights of the edges in the store interaction graph of the updated features; according to the weights of the edges in the store interaction graph of the updated features, the features of the nodes in the store interaction graph of the updated features are updated again to obtain the second features of the sample object, the second features of the target sample commodity, and the second features of the target sample store.

[0061] Optionally, the initial features of the sample object node, the target sample store node, and the target sample store node are irrelevant to the interactive behavior. Therefore, based on the connection relationship of the nodes and edges of the store interaction graph and the features of the adjacent nodes, the initial features of the sample object node, the target sample store node, and the target sample store node are updated so that the updated initial features are related to the interactive behavior. Since the sample object has different degrees of interest in different target store sample nodes, the weight of the corresponding edge in the store interaction graph is calculated based on the updated features of the target sample store node, and then the second feature of the sample object is calculated based on the weight of the edge in the obtained store interaction graph.

[0062] Specifically, based on the store interaction graph G S , according to the characteristics of the nodes adjacent to the sample object node, the initial characteristics of the sample object Update, and, according to the characteristics of the nodes adjacent to the target sample product node, update the initial characteristics of the target sample product Update, and, according to the characteristics of the nodes adjacent to the target sample store node, update the initial characteristics Vs of the target sample store S Update, expressed as

[0063]

[0064] in, is the initial feature of the updated sample object, V s s' is the initial feature of the updated target sample store, V i s' is the initial feature of the updated target sample product. Based on the updated features of the target sample store node, the second feature of the sample object is calculated, which is expressed as:

[0065]

[0066] in, is the first feature of the sample object.

[0067] It should be noted that in the time period T n When based on the store interaction diagram According to the features of the nodes adjacent to the sample object node, the initial features of the sample object are Update, and, according to the characteristics of the nodes adjacent to the target sample product node, update the initial characteristics of the target sample product Update, and, according to the characteristics of the nodes adjacent to the target sample store node, update the initial characteristics of the target sample store Update, expressed as

[0068]

[0069] in, is the updated sample object in time period T n The initial characteristics of is the updated target sample store in time period T n The initial characteristics of is the updated target sample product in time period T n Based on the updated features of the target sample store node, calculate the sample object in time period T n The first feature of is expressed as:

[0070]

[0071] in, is the sample object in time period T n The first feature.

[0072] As an optional embodiment, the original neural network model is trained based on positive samples to obtain a neural network model, including: determining other sample products that have no interactive behavior with the sample object; generating negative samples based on the initial features of other sample products and the features of the sample object; and training the original neural network model based on the positive samples and the negative samples to obtain a neural network model.

[0073] Optionally, since the sample object does not interact with all sample products, the samples generated when the sample object interacts with the sample product are classified as positive samples, and the samples generated with other sample products without interaction are classified as negative samples. The positive samples and negative samples together constitute training samples, which are used to train the original neural network model to obtain a neural network model.

[0074] Figure 3 It is a schematic diagram of the object feature and product feature extraction process provided according to an optional embodiment of the present invention. Obtain the object feature of the target object and the product features of multiple products, wherein the product features of multiple products include the features of the store to which the products belong; combine the object feature with the product features of multiple products to obtain multiple sets of input data; input the multiple sets of input data into a pre-trained neural network model, and the neural network model sorts the multiple sets of input data, wherein the neural network model is obtained by training the original neural network model with multiple training samples, each of the multiple training samples includes the features of the sample object and the features of the sample product, and the multiple training samples include positive samples with interactive behaviors, and the features of the sample objects included in the positive samples are determined by the interactive behaviors between the sample object, the sample product and the sample store; recommend products to the target object based on the sorted multiple sets of input data.

[0075] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0076] Through the description of the above implementation methods, those skilled in the art can clearly understand that the commodity recommendation method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0077] According to an embodiment of the present invention, a commodity recommendation device for implementing the above commodity recommendation method is also provided. Figure 4 is a structural block diagram of a commodity recommendation device provided according to an embodiment of the present invention. Figure 4 As shown, the commodity recommendation device includes: an acquisition module 41, a combination module 42, a training module 43, and a recommendation module 44. The commodity recommendation device is described below.

[0078] An acquisition module, used to acquire object features of a target object and product features of each of a plurality of products, wherein the product features of each of the plurality of products include features of a store to which the products belong;

[0079] A combining module, used for combining the object feature with the commodity features of the plurality of commodities to obtain a plurality of groups of input data;

[0080] A training module, used for inputting the multiple groups of input data into a pre-trained neural network model, and allowing the neural network model to sort the multiple groups of input data, wherein the neural network model is obtained by training an original neural network model through multiple training samples, each of the multiple training samples includes features of a sample object and features of a sample commodity, and the multiple training samples include positive samples with interactive behaviors, and features of the sample objects included in the positive samples are determined by interactive behaviors among the sample objects, the sample commodities, and the sample stores;

[0081] The recommendation module is used to recommend commodities to the target object according to the sorted multiple groups of input data.

[0082] It should be noted that the acquisition module 41, the combination module 42, the training module 43 and the recommendation module 44 correspond to steps S202 to S208 in the embodiment, and the examples and application scenarios implemented by the multiple modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above embodiment. It should be noted that the above modules as part of the device can be run in the computer terminal 10 provided in the embodiment.

[0083] An embodiment of the present invention may provide a computer device. Optionally, in this embodiment, the computer device may be located in at least one network device among multiple network devices of a computer network. The computer device includes a memory and a processor.

[0084] Among them, the memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the commodity recommendation method and device in the embodiment of the present invention. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, realizing the above-mentioned commodity recommendation method. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories can be connected to the computer terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0085] The processor can call the information and application programs stored in the memory through the transmission device to execute the following steps: obtain the object characteristics of the target object and the product characteristics of each of the multiple commodities, wherein the product characteristics of each of the multiple commodities include the characteristics of the store to which the commodities belong; combine the object characteristics with the product characteristics of each of the multiple commodities to obtain multiple groups of input data; input the multiple groups of input data into a pre-trained neural network model, and the neural network model sorts the multiple groups of input data, wherein the neural network model is obtained by training the original neural network model through multiple training samples, each of the multiple training samples includes the characteristics of the sample object and the characteristics of the sample commodity, and the multiple training samples include positive samples with interactive behaviors, and the characteristics of the sample object included in the positive samples are determined by the interactive behaviors between the sample object, the sample commodity and the sample store; and recommend commodities to the target object based on the sorted multiple groups of input data.

[0086] Optionally, the neural network model is trained in the following manner: obtaining initial features of the sample object and initial features of the sample product; determining a target sample product that interacts with the sample object; updating the initial features of the sample object based on the interaction between the sample object and the target sample product to obtain features of the sample object, and updating the initial features of the target sample product to obtain features of the target sample product; generating positive samples based on the features of the sample object and the features of the target sample product; and training the original neural network model based on the positive samples to obtain a neural network model.

[0087] Optionally, according to the interaction behavior between the sample object and the target sample product, the initial features of the sample object are updated to obtain the features of the sample object, and the initial features of the target sample product are updated to obtain the features of the target sample product, including: obtaining the initial features of the target sample store that interacts with the sample object; constructing a product interaction graph and a store interaction graph according to the interaction behavior between the sample object and the target sample product, and the interaction behavior between the sample object and the target sample store, wherein the product interaction graph includes sample object nodes, target sample product nodes, target sample store nodes, edges between the sample object nodes and the target sample product nodes, and edges between the target sample product nodes and the target sample store nodes, and the store interaction graph includes sample object nodes, target sample product nodes, target sample store nodes, edges between the sample object nodes and the target sample store nodes, and edges between the target sample product nodes and the target sample store nodes; based on the product interaction graph, the sample object nodes are updated. The initial characteristics of the sample object of the point are updated to obtain the first characteristics of the sample object, and the initial characteristics of the target sample commodity of the target sample commodity node are updated to obtain the first characteristics of the target sample commodity, and the initial characteristics of the target sample store of the target sample store node are updated to obtain the first characteristics of the target sample store; based on the store interaction graph, the initial characteristics of the sample object of the sample object node are updated to obtain the second characteristics of the sample object, and the initial characteristics of the target sample commodity of the target sample commodity node are updated to obtain the second characteristics of the target sample commodity, and the initial characteristics of the target sample store of the target sample store node are updated to obtain the second characteristics of the target sample store; according to the first characteristics of the sample object and the second characteristics of the sample object, the characteristics of the sample object are obtained, and according to the first characteristics of the target sample commodity, the second characteristics of the target sample commodity, the first characteristics of the target sample store and the second characteristics of the target sample store, the characteristics of the target sample commodity are obtained.

[0088] Optionally, based on the commodity interaction graph, the initial features of the sample object of the sample object node are updated to obtain the first feature of the sample object, and the initial features of the target sample commodity of the target sample commodity node are updated to obtain the first feature of the target sample commodity, and the initial features of the target sample store of the target sample store node are updated to obtain the first feature of the target sample store, including: based on the commodity interaction graph, the initial features of the sample object are updated according to the features of the nodes adjacent to the sample object node, and the initial features of the target sample commodity are updated according to the features of the nodes adjacent to the target sample commodity node, and the initial features of the target sample store are updated according to the features of the nodes adjacent to the target sample store node to obtain the commodity interaction graph of the updated features; attention extraction is performed on the commodity interaction graph of the updated features to determine the weights of the edges in the commodity interaction graph of the updated features; according to the weights of the edges in the commodity interaction graph of the updated features, the features of the nodes in the commodity interaction graph of the updated features are updated again to obtain the first feature of the sample object, the first feature of the target sample commodity, and the first feature of the target sample store.

[0089] Optionally, based on the store interaction graph, the initial features of the sample object of the sample object node are updated to obtain the second features of the sample object, and the initial features of the target sample commodity of the target sample commodity node are updated to obtain the second features of the target sample commodity, and the initial features of the target sample shop of the target sample shop node are updated to obtain the second features of the target sample shop, including: based on the store interaction graph, the initial features of the sample object are updated according to the features of the nodes adjacent to the sample object node, and the initial features of the target sample commodity are updated according to the features of the nodes adjacent to the target sample commodity node, and the initial features of the target sample shop are updated according to the features of the nodes adjacent to the target sample shop node to obtain the store interaction graph of the updated features; attention extraction is performed on the store interaction graph of the updated features to determine the weights of the edges in the store interaction graph of the updated features; according to the weights of the edges in the store interaction graph of the updated features, the features of the nodes in the store interaction graph of the updated features are updated again to obtain the second features of the sample object, the second features of the target sample commodity, and the second features of the target sample shop.

[0090] Optionally, based on the positive samples, the original neural network model is trained to obtain the neural network model, including: determining other sample products that have no interactive behavior with the sample object; generating negative samples based on the initial features of the other sample products and the features of the sample object; based on the positive samples and the negative samples, the original neural network model is trained to obtain the neural network model.

[0091] By adopting the embodiment of the present invention, a solution for product recommendation is provided, which achieves the purpose of deeply mining the characteristics of products that users are interested in, thereby realizing the technical effect of accurately and efficiently recommending products of interest to users, and further solves the technical problem that traditional methods are unable to deeply mine the characteristics of products that users are interested in, resulting in low product recommendation efficiency.

[0092] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a non-volatile storage medium, and the storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0093] The embodiment of the present invention further provides a non-volatile storage medium. Optionally, in this embodiment, the non-volatile storage medium can be used to store the program code executed by the commodity recommendation method provided in the embodiment.

[0094] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.

[0095] Optionally, in this embodiment, the non-volatile storage medium is configured to store program codes for executing the following steps: obtaining object features of a target object and product features of multiple commodities, wherein the product features of the multiple commodities include features of the store to which the commodities belong; combining the object features with the product features of the multiple commodities to obtain multiple groups of input data; inputting the multiple groups of input data into a pre-trained neural network model, and sorting the multiple groups of input data by the neural network model, wherein the neural network model is obtained by training an original neural network model with multiple training samples, each of the multiple training samples includes features of a sample object and features of a sample commodity, and the multiple training samples include positive samples with interactive behaviors, and features of the sample objects included in the positive samples are determined by the interactive behaviors among the sample object, the sample commodity and the sample store; and recommending commodities to the target object based on the sorted multiple groups of input data.

[0096] Optionally, the neural network model is trained in the following manner: obtaining initial features of the sample object and initial features of the sample product; determining a target sample product that interacts with the sample object; updating the initial features of the sample object based on the interaction between the sample object and the target sample product to obtain features of the sample object, and updating the initial features of the target sample product to obtain features of the target sample product; generating positive samples based on the features of the sample object and the features of the target sample product; and training the original neural network model based on the positive samples to obtain a neural network model.

[0097] Optionally, according to the interaction behavior between the sample object and the target sample product, the initial features of the sample object are updated to obtain the features of the sample object, and the initial features of the target sample product are updated to obtain the features of the target sample product, including: obtaining the initial features of the target sample store that interacts with the sample object; constructing a product interaction graph and a store interaction graph according to the interaction behavior between the sample object and the target sample product, and the interaction behavior between the sample object and the target sample store, wherein the product interaction graph includes sample object nodes, target sample product nodes, target sample store nodes, edges between the sample object nodes and the target sample product nodes, and edges between the target sample product nodes and the target sample store nodes, and the store interaction graph includes sample object nodes, target sample product nodes, target sample store nodes, edges between the sample object nodes and the target sample store nodes, and edges between the target sample product nodes and the target sample store nodes; based on the product interaction graph, the sample object nodes are updated. The initial characteristics of the sample object of the point are updated to obtain the first characteristics of the sample object, and the initial characteristics of the target sample commodity of the target sample commodity node are updated to obtain the first characteristics of the target sample commodity, and the initial characteristics of the target sample store of the target sample store node are updated to obtain the first characteristics of the target sample store; based on the store interaction graph, the initial characteristics of the sample object of the sample object node are updated to obtain the second characteristics of the sample object, and the initial characteristics of the target sample commodity of the target sample commodity node are updated to obtain the second characteristics of the target sample commodity, and the initial characteristics of the target sample store of the target sample store node are updated to obtain the second characteristics of the target sample store; according to the first characteristics of the sample object and the second characteristics of the sample object, the characteristics of the sample object are obtained, and according to the first characteristics of the target sample commodity, the second characteristics of the target sample commodity, the first characteristics of the target sample store and the second characteristics of the target sample store, the characteristics of the target sample commodity are obtained.

[0098] Optionally, based on the commodity interaction graph, the initial features of the sample object of the sample object node are updated to obtain the first feature of the sample object, and the initial features of the target sample commodity of the target sample commodity node are updated to obtain the first feature of the target sample commodity, and the initial features of the target sample store of the target sample store node are updated to obtain the first feature of the target sample store, including: based on the commodity interaction graph, the initial features of the sample object are updated according to the features of the nodes adjacent to the sample object node, and the initial features of the target sample commodity are updated according to the features of the nodes adjacent to the target sample commodity node, and the initial features of the target sample store are updated according to the features of the nodes adjacent to the target sample store node to obtain the commodity interaction graph of the updated features; attention extraction is performed on the commodity interaction graph of the updated features to determine the weights of the edges in the commodity interaction graph of the updated features; according to the weights of the edges in the commodity interaction graph of the updated features, the features of the nodes in the commodity interaction graph of the updated features are updated again to obtain the first feature of the sample object, the first feature of the target sample commodity, and the first feature of the target sample store.

[0099] Optionally, based on the store interaction graph, the initial features of the sample object of the sample object node are updated to obtain the second features of the sample object, and the initial features of the target sample commodity of the target sample commodity node are updated to obtain the second features of the target sample commodity, and the initial features of the target sample shop of the target sample shop node are updated to obtain the second features of the target sample shop, including: based on the store interaction graph, the initial features of the sample object are updated according to the features of the nodes adjacent to the sample object node, and the initial features of the target sample commodity are updated according to the features of the nodes adjacent to the target sample commodity node, and the initial features of the target sample shop are updated according to the features of the nodes adjacent to the target sample shop node to obtain the store interaction graph of the updated features; attention extraction is performed on the store interaction graph of the updated features to determine the weights of the edges in the store interaction graph of the updated features; according to the weights of the edges in the store interaction graph of the updated features, the features of the nodes in the store interaction graph of the updated features are updated again to obtain the second features of the sample object, the second features of the target sample commodity, and the second features of the target sample shop.

[0100] Optionally, based on the positive samples, the original neural network model is trained to obtain the neural network model, including: determining other sample products that have no interactive behavior with the sample object; generating negative samples based on the initial features of the other sample products and the features of the sample object; based on the positive samples and the negative samples, the original neural network model is trained to obtain the neural network model.

[0101] An embodiment of the present invention further provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the product recommendation method in each embodiment of the present application.

[0102] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0103] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0104] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0105] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0106] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0107] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a non-volatile storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0108] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A product recommendation method, characterized in that: include: Acquire object features of the target object and product features of each of the plurality of products, wherein the product features of each of the plurality of products include features of the store to which the product belongs; Combining the object features with the commodity features of the plurality of commodities respectively to obtain a plurality of groups of input data; Inputting the multiple groups of input data into a pre-trained neural network model, and using the neural network model to sort the multiple groups of input data, wherein the neural network model is obtained by training an original neural network model through multiple training samples, each of the multiple training samples includes features of a sample object and features of a sample commodity, and the multiple training samples include positive samples with interactive behaviors, and features of the sample objects included in the positive samples are determined by interactive behaviors among the sample objects, the sample commodities, and the sample stores; Recommend commodities to the target object according to the sorted multiple groups of input data.

2. The method according to claim 1, characterized in that The neural network model is trained in the following way: Acquiring initial features of the sample object and initial features of the sample commodity; Determining target sample commodities that have interactive behaviors with the sample object; According to the interaction behavior between the sample object and the target sample commodity, the initial features of the sample object are updated to obtain the features of the sample object, and the initial features of the target sample commodity are updated to obtain the features of the target sample commodity; Generating the positive sample according to the characteristics of the sample object and the characteristics of the target sample commodity; Based on the positive samples, the original neural network model is trained to obtain the neural network model.

3. The method according to claim 2, characterized in that The updating of the initial features of the sample object according to the interaction between the sample object and the target sample commodity to obtain the features of the sample object, and the updating of the initial features of the target sample commodity to obtain the features of the target sample commodity include: Acquire initial features of a target sample store that interacts with the sample object; According to the interaction behavior between the sample object and the target sample product, and the interaction behavior between the sample object and the target sample store, a product interaction graph and a store interaction graph are constructed, wherein the product interaction graph includes a sample object node, a target sample product node, a target sample store node, edges between the sample object node and the target sample product node, and edges between the target sample product node and the target sample store node, and the store interaction graph includes the sample object node, the target sample product node, the target sample store node, edges between the sample object node and the target sample store node, and edges between the target sample product node and the target sample store node; Based on the commodity interaction graph, the initial feature of the sample object of the sample object node is updated to obtain the first feature of the sample object, and the initial feature of the target sample commodity of the target sample commodity node is updated to obtain the first feature of the target sample commodity, and the initial feature of the target sample store of the target sample store node is updated to obtain the first feature of the target sample store; Based on the store interaction graph, the initial feature of the sample object of the sample object node is updated to obtain the second feature of the sample object, and the initial feature of the target sample commodity of the target sample commodity node is updated to obtain the second feature of the target sample commodity, and the initial feature of the target sample store of the target sample store node is updated to obtain the second feature of the target sample store; The characteristics of the sample object are obtained according to the first characteristics of the sample object and the second characteristics of the sample object, and the characteristics of the target sample product are obtained according to the first characteristics of the target sample product, the second characteristics of the target sample product, the first characteristics of the target sample store and the second characteristics of the target sample store.

4. The method according to claim 3, characterized in that The method of updating the initial feature of the sample object of the sample object node based on the commodity interaction graph to obtain the first feature of the sample object, updating the initial feature of the target sample commodity of the target sample commodity node to obtain the first feature of the target sample commodity, and updating the initial feature of the target sample store of the target sample store node to obtain the first feature of the target sample store includes: Based on the commodity interaction graph, the initial features of the sample object are updated according to the features of the nodes adjacent to the sample object node, and the initial features of the target sample commodity are updated according to the features of the nodes adjacent to the target sample commodity node, and the initial features of the target sample store are updated according to the features of the nodes adjacent to the target sample store node, to obtain a commodity interaction graph with updated features; Performing attention extraction on the commodity interaction graph of the updated features, and determining the weights of edges in the commodity interaction graph of the updated features; According to the weights of the edges in the commodity interaction graph of the updated features, the features of the nodes in the commodity interaction graph of the updated features are updated again to obtain the first feature of the sample object, the first feature of the target sample commodity, and the first feature of the target sample store.

5. The method according to claim 3, characterized in that: The method of updating the initial feature of the sample object of the sample object node based on the store interaction graph to obtain the second feature of the sample object, updating the initial feature of the target sample commodity of the target sample commodity node to obtain the second feature of the target sample commodity, and updating the initial feature of the target sample store of the target sample store node to obtain the second feature of the target sample store includes: Based on the store interaction graph, the initial features of the sample object are updated according to the features of the nodes adjacent to the sample object node, and the initial features of the target sample commodity are updated according to the features of the nodes adjacent to the target sample commodity node, and the initial features of the target sample store are updated according to the features of the nodes adjacent to the target sample store node, to obtain a store interaction graph with updated features; Performing attention extraction on the store interaction graph of the updated features, and determining the weights of edges in the store interaction graph of the updated features; According to the weights of the edges in the store interaction graph of the updated features, the features of the nodes in the store interaction graph of the updated features are updated again to obtain the second feature of the sample object, the second feature of the target sample product, and the second feature of the target sample store.

6. The method according to claim 2, characterized in that The training of the original neural network model based on the positive sample to obtain the neural network model includes: Determine other sample products that have no interaction with the sample object; Generate a negative sample according to the initial features of the other sample products and the features of the sample object; Based on the positive samples and the negative samples, the original neural network model is trained to obtain the neural network model.

7. A commodity recommendation device, characterized in that: include: An acquisition module, used to acquire object features of a target object and product features of each of a plurality of products, wherein the product features of each of the plurality of products include features of a store to which the products belong; A combining module, used for combining the object feature with the commodity features of the plurality of commodities to obtain a plurality of groups of input data; A training module, used for inputting the multiple groups of input data into a pre-trained neural network model, and allowing the neural network model to sort the multiple groups of input data, wherein the neural network model is obtained by training an original neural network model through multiple training samples, each of the multiple training samples includes features of a sample object and features of a sample commodity, and the multiple training samples include positive samples with interactive behaviors, and features of the sample objects included in the positive samples are determined by interactive behaviors among the sample objects, the sample commodities, and the sample stores; The recommendation module is used to recommend commodities to the target object according to the sorted multiple groups of input data.

8. A non-volatile storage medium, characterized in that: The non-volatile storage medium includes a stored program, wherein when the program is executed, the device where the non-volatile storage medium is located is controlled to execute the commodity recommendation method according to any one of claims 1 to 6.

9. A computer device, characterized in that: include: Memory and processor, The memory stores a computer program; The processor is used to execute the computer program stored in the memory, and when the computer program is running, the processor executes the product recommendation method according to any one of claims 1 to 6.

10. A computer program product comprising computer instructions, characterized in that: The computer instructions are executed by a processor to execute the product recommendation method described in any one of claims 1 to 6.