Intelligent product selection management method and system based on machine learning

Abstract

The embodiment of the invention belongs to the field of artificial intelligence, and relates to an intelligent article selection management method based on machine learning, which comprises the steps of generating a customer preference portrait of each customer based on historical purchase data and browsing data; carrying out attribute embedding and knowledge graph embedding on the product information of each product to obtain a product portrait of each product; constructing a customer product interaction diagram according to the historical purchase data and the browsing data; performing representation learning on the customer product interaction diagram to obtain product node characteristics of each product node; according to the product portrait of each product and the product node feature of the corresponding product node, generating a product comprehensive feature of each product; the customer preference portraits of the target customers and the product comprehensive features of the products are combined to obtain combined features, the combined features are input into a matching model, and interestingness between the target customers and the products is obtained; and determining a target product in the products according to the obtained interestingness to perform product recommendation. According to the invention, the accuracy of product selection and recommendation is improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to an intelligent product selection management method, device, computer equipment and storage medium based on machine learning. Background Art

[0002] In e-commerce platforms, there are a large number of products with a wide variety of types. Consumers often face information overload and decision-making difficulties when choosing products. At the same time, the purchasing preferences and needs of different consumers vary greatly. Traditional product recommendation technologies face many technical challenges. How to accurately capture and understand consumer needs is a major problem. Secondly, the attribute characteristics of products have many dimensions, and there may be complex correlations between different attributes. How to establish a comprehensive and accurate product portrait is another challenge. Furthermore, in the context of massive product and user data, how to efficiently build a matching relationship between users and products also puts forward high requirements. Therefore, there is an urgent need for an intelligent product selection and recommendation method to improve the accuracy of product selection and recommendation. Summary of the invention

[0003] The purpose of the embodiments of the present application is to propose an intelligent product selection management method, device, computer equipment and storage medium based on machine learning to solve the accuracy of product selection and recommendation. In order to solve the above technical problems, the embodiment of the present application provides an intelligent product selection management method based on machine learning, which adopts the following technical solutions: Based on the historical purchase data and browsing data of each customer, a customer preference profile of each customer is generated by a preset profile generation algorithm; Obtain product information of each product, and embed attributes and knowledge graphs of the product information of each product to obtain product portraits of each product; Constructing a customer-product interaction graph based on the obtained historical purchase data and browsing data, wherein the customer-product interaction graph includes customer nodes and product nodes; Performing representation learning on the customer-product interaction graph to obtain product node features of each product node; Generate comprehensive product features of each product according to the product profile of each product and the product node features of its corresponding product node; Combining the customer preference profile of the target customer and the comprehensive product features of each product to obtain combined features, and inputting each combined feature into a matching model to obtain the interest degree between the target customer and each product; A target product is determined from among the products according to the obtained interest level, and a product recommendation is made to the target customer according to the determined target product.

[0004] Furthermore, the step of generating the customer preference portrait of each customer by a preset portrait generation algorithm based on the historical purchase data and browsing data of each customer includes: Based on the historical purchase data of each customer, the customer similarity between the customers is calculated by a collaborative filtering algorithm, and similar customers of the target customer are determined from among the customers according to the obtained customer similarity, wherein the target customer comes from the customers; Determine the purchase preference products of each similar customer according to the historical purchase data of each similar customer, and construct the first product set of the target customer according to each purchase preference product; Based on the browsing data of each customer, calculating the product similarity between each product by a content-based recommendation algorithm; Determine browsing preference products similar to the current browsing product of the target customer according to the obtained product similarity, and add each browsing preference product to the first product set to obtain a second product set of the target customer; A customer preference profile of the target customer is generated based on the second product set, historical purchase data and browsing data of the target customer, and a customer preference profile of each customer is obtained.

[0005] Furthermore, after the step of obtaining the second product set of the target customer, the method further includes: Process the historical purchase data of each customer through an association rule mining algorithm to obtain product purchase association information; According to the product purchase association information, purchase-associated products are added to the second product set.

[0006] Furthermore, the step of embedding attributes and knowledge graphs into the product information of each product to obtain a product portrait of each product includes: Preprocessing the unstructured data in the product information of each product to obtain the structured data of each product; Extracting attribute information from the structured data of each product to obtain structured attribute data of each product; Mapping the structured attribute data of each product into attribute embedding representations respectively; Based on the attribute embedding representation of each product, determine the association relationship between the products through similarity calculation and clustering algorithm, and construct a product knowledge graph according to the obtained association relationship; The attribute embedding representation of each product and the knowledge graph embedding representation are fused to obtain the product portrait of each product.

[0007] Furthermore, the step of performing representation learning on the customer-product interaction graph to obtain product node features of each product node includes: Generate a node sequence on the customer-product interaction graph by a random walk algorithm; The node sequence is input into a skip-gram model to generate product node features of each product node in the customer-product interaction graph by maximizing the node co-occurrence probability.

[0008] Furthermore, the method further comprises: For each product node in the customer-product interaction graph, determining a neighbor node of the product node; Calculating the node similarity between the product node and each neighbor node; Determine the attention weight of each neighbor node according to the obtained node similarity; Aggregating the product node features of each neighbor node according to the obtained attention weights to obtain aggregated neighbor information; The product node feature of the product node is updated according to the aggregated neighbor information.

[0009] Furthermore, the step of generating the comprehensive product features of each product according to the product portrait of each product and the product node features of its corresponding product node includes: For each product, extract a product embedding vector of the product from the product portrait of the product; The product embedding vector is combined with the product node feature of the product node corresponding to the product to obtain the comprehensive product feature of the product.

[0010] In order to solve the above technical problems, the embodiment of the present application also provides an intelligent product selection management device based on machine learning, which adopts the following technical solution: A preference generation module, used to generate a customer preference profile of each customer through a preset profile generation algorithm based on the historical purchase data and browsing data of each customer; A product generation module is used to obtain product information of each product, and embed attributes and knowledge graphs into the product information of each product to obtain a product profile of each product; An interaction construction module, used to construct a customer-product interaction graph based on the obtained historical purchase data and browsing data, wherein the customer-product interaction graph includes customer nodes and product nodes; A feature generation module, used for performing representation learning on the customer-product interaction graph to obtain product node features of each product node; A comprehensive generation module, used to generate comprehensive product features of each product according to the product profile of each product and the product node features of its corresponding product node; A combination calculation module, used to combine the customer preference profile of the target customer and the comprehensive product features of each product to obtain a combination feature, and input each combination feature into a matching model to obtain the interest degree between the target customer and each product; The product recommendation module is used to determine a target product from among the products according to the obtained interest level, and to make product recommendations to the target customer according to the determined target product.

[0011] In order to solve the above technical problems, an embodiment of the present application also provides a computer device, which includes a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, it implements the steps of the above-mentioned intelligent product selection management method based on machine learning.

[0012] In order to solve the above-mentioned technical problems, an embodiment of the present application also provides a computer-readable storage medium, on which computer-readable instructions are stored. When the computer-readable instructions are executed by a processor, the steps of the above-mentioned intelligent product selection management method based on machine learning are implemented.

[0013] Compared with the prior art, the embodiments of the present application mainly have the following beneficial effects: generating a customer preference portrait of each customer through historical purchase data and browsing data, embedding attributes and knowledge graphs into product information of each product, and generating high-quality product portraits; constructing a customer-product interaction graph and performing representation learning, further enhancing the implicit correlation representation between product nodes; combining the product portraits of each product and the product node features of its corresponding product nodes, and converting the complex relationship between customers and products into comprehensive product features represented in multiple dimensions; inputting customer preference portraits and comprehensive product features into a matching model to accurately calculate interest, thereby achieving personalized recommendations, improving the accuracy of product selection and recommendations, and improving the conversion rate of product recommendations and customer satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0015] Figure 1 is an exemplary system architecture diagram to which the present application may be applied; Figure 2 is a flowchart of an embodiment of an intelligent product selection management method based on machine learning according to the present application; Figure 3It is a structural schematic diagram of an embodiment of an intelligent product selection management device based on machine learning according to the present application; Figure 4 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0016] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0017] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0018] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0019] like Figure 1 As shown, the system architecture 100 may include a terminal device 101, a network 102 and a server 103. The terminal device 101 may be a laptop 1011, a tablet computer 1012 or a mobile phone 1013. The network 102 is used to provide a medium for a communication link between the terminal device 101 and the server 103. The network 102 may include various connection types, such as wired, wireless communication links or optical fiber cables.

[0020] The user can use the terminal device 101 to interact with the server 103 through the network 102 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0021] The terminal device 101 can be any electronic device with a display screen and supporting web browsing. In addition to the laptop computer 1011, the tablet computer 1012 or the mobile phone 1013, the terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 (Moving Picture Experts Group Audio Layer IV), a laptop computer, a desktop computer, etc.

[0022] The server 103 may be a server that provides various services, such as a background server that provides support for a web page displayed on the terminal device 101 .

[0023] It should be noted that the intelligent product selection management method based on machine learning provided in the embodiment of the present application is generally executed by a server. Accordingly, an intelligent product selection management device based on machine learning is generally arranged in the server.

[0024] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to the implementation requirements.

[0025] Continue to refer Figure 2 , shows a flow chart of an embodiment of an intelligent product selection management method based on machine learning according to the present application. The intelligent product selection management method based on machine learning comprises the following steps: Step S201, based on the historical purchase data and browsing data of each customer, a customer preference profile of each customer is generated by a preset profile generation algorithm.

[0026] In this embodiment, an intelligent product selection management method based on machine learning is run on an electronic device (e.g. Figure 1 The server shown in the figure) can communicate with the terminal device through a wired connection or a wireless connection. It should be noted that the above wireless connection method may include but is not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future.

[0027] Specifically, the historical purchase data and browsing data of each customer are obtained. The historical purchase data can be the customer's past purchase records on the platform, and the browsing data can be the customer's behavior data of browsing products on the platform. These data can reflect the customer's preferences for different product categories, price ranges, brands, etc.

[0028] The historical purchase data and browsing data can be processed step by step according to the preset profile generation algorithm to generate customer preference profiles for each customer. Customer preference profiles are generated by analyzing the customer's historical purchase and browsing behaviors, aiming to capture the customer's consumption habits, interests and needs. The profile generation algorithm can include steps such as data cleaning, feature extraction, and behavior analysis to convert historical data into a digital representation of customer preferences.

[0029] Step S202, obtain product information of each product, and embed attributes and knowledge graphs into the product information of each product to obtain a product portrait of each product.

[0030] Specifically, product portraits are characteristic information that describes a product, which may include structured data such as product attributes, categories, brands, prices, and unstructured data such as user-generated content (such as comments, tags, etc.).

[0031] The generation of product portraits is achieved through attribute embedding and knowledge graph embedding. Attribute embedding maps product attribute data (such as price, brand, etc.) into a low-dimensional space to obtain its feature representation; knowledge graph embedding constructs a knowledge graph by utilizing the associations between products (for example, similar products, complementary products, etc.) to further enhance the representation of product features. In the knowledge graph, nodes can be products and edges can be the relationships between products. The associations between product attribute data and products come from product information.

[0032] Step S203, constructing a customer-product interaction graph based on the obtained historical purchase data and browsing data, wherein the customer-product interaction graph includes customer nodes and product nodes.

[0033] Specifically, based on each customer's historical purchase data and browsing data, a customer-product interaction graph can be constructed. The customer-product interaction graph describes the relationship between customers and products. In the graph, nodes represent customers or products, and edges represent interactions between customers and products (such as purchases, browsing, etc.). The graph structure helps discover the implicit potential relationship between customers and products.

[0034] The customer-product interaction graph is an interaction network between users and products. For example, if a customer purchases a product, the customer node and the product node are connected in the customer-product interaction graph to form an edge.

[0035] Step S204, performing representation learning on the customer-product interaction graph to obtain product node features of each product node.

[0036] Specifically, representation learning is to train each node (here, the product node) in the customer-product interaction graph to have a low-dimensional vector representation, which can fully express the characteristics of the node and the relationship between the nodes. Graph representation learning methods can include DeepWalk, node2vec, etc.

[0037] The vector representing the product node is called the product node feature. The product node feature of each product node is an embedded vector obtained through graph learning methods. These feature vectors can capture the relationship and similarity with other product nodes.

[0038] Step S205, generating comprehensive product features for each product based on the product portraits of each product and the product node features of its corresponding product nodes.

[0039] Specifically, the comprehensive product features combine the product attributes (such as brand, price, type, etc., which can come from the product portrait) with the product node features obtained by graph representation learning to form a multi-dimensional feature representation of the product. This feature representation can reflect the basic attributes of the product and combine the similarity or relationship with other products.

[0040] Step S206, combining the customer preference profile of the target customer and the comprehensive product features of each product to obtain combined features, and inputting each combined feature into a matching model to obtain the interest level between the target customer and each product.

[0041] Specifically, the customer preference profile of the target customer and the comprehensive product features of each product are combined (e.g., concatenated, weighted, etc.) to form a complete representation to fully reflect the relationship between the target customer and each product. The combined features obtained after the combination are input into a trained matching model (e.g., a neural network model, a support vector machine, etc.), and the interest between the target customer and each product is calculated through the matching model. The higher the interest, the more likely the customer is to be interested in the product.

[0042] Step S207, determining a target product from among the products according to the obtained interest level, and recommending products to the target customers according to the determined target product.

[0043] Specifically, based on the target customers' interest in each product, select products with higher interest as target products. These target products may satisfy the potential needs of customers. Recommending products to target customers based on target products can help improve the customer's purchase conversion rate and improve the accuracy of product selection and recommendation.

[0044] In this embodiment, a customer preference portrait of each customer is generated through historical purchase data and browsing data, and the product information of each product is embedded in attributes and knowledge graphs to generate high-quality product portraits; a customer-product interaction graph is constructed and representation learning is performed to further enhance the implicit correlation representation between product nodes; the product portrait of each product and the product node features of its corresponding product nodes are combined to convert the complex relationship between customers and products into comprehensive product features represented in multiple dimensions; the customer preference portrait and the comprehensive product features are input into a matching model to accurately calculate the interest level, thereby achieving personalized recommendations, improving the accuracy of product selection and recommendations, and improving the conversion rate of product recommendations and customer satisfaction.

[0045] Furthermore, the above step S201 may include: based on the historical purchase data of each customer, calculating the customer similarity between each customer through a collaborative filtering algorithm, and determining similar customers of the target customer among each customer according to the obtained customer similarity, wherein the target customer comes from each customer; determining the purchase preference products of each similar customer according to the historical purchase data of each similar customer, and constructing a first product set for the target customer according to each purchase preference product; based on the browsing data of each customer, calculating the product similarity between each product through a content-based recommendation algorithm; determining the browsing preference products similar to the current browsing products of the target customer according to the obtained product similarity, and adding each browsing preference product to the first product set to obtain a second product set for the target customer; generating a customer preference portrait of the target customer according to the second product set, historical purchase data and browsing data of the target customer, and obtaining a customer preference portrait of each customer.

[0046] Specifically, through the collaborative filtering algorithm, the customer similarity between the target customer and other customers is calculated based on the historical purchase data of the target customer and other customers. The calculation of customer similarity can adopt user-based collaborative filtering to find similar customers by comparing purchase behaviors (such as purchased products, purchase frequency, etc.).

[0047] Collaborative filtering is a recommendation algorithm that makes recommendations by analyzing the similarities between users or items. The user-based collaborative filtering method predicts the products that a user may be interested in based on the similarity between the user and other users.

[0048] Based on the calculated customer similarity, find other customers with the most similar purchasing behavior to the target customer. The purchasing preferences of these similar customers can recommend suitable products to the target customer. Extract the products that similar customers prefer to buy from their purchase history data, i.e., purchase preference products, and determine the extracted purchase preference products as the first product set for the target customer. Purchase preference products are inferred based on the purchase preferences of similar customers and can reflect the potential interests of the target customer.

[0049] Based on the target customer's browsing data, a content-based recommendation algorithm is used to calculate the similarity between the customer's current browsing product and other products. This step focuses on the attributes of the product itself, such as category, brand, price, etc. Based on product similarity, products with higher similarity are selected from other products similar to the target customer's current browsing product as browsing preference products, and added to the first product set to obtain the second product set. These products reflect the target customer's current interests and needs.

[0050] The target customer's historical purchase data, browsing data and second product set are integrated to generate a customer preference profile of the target customer. It describes the target customer's interest level and purchase tendency for different categories of products. In one embodiment, the customer preference profile may also include customer information of the target customer.

[0051] In this embodiment, by combining historical purchase data and browsing data, recommendations can be generated based on the customer's past behavior, and the recommended content can be adjusted according to the current browsing preferences, thereby expanding the diversity of recommendations and facilitating more personalized and accurate product recommendations; real-time browsing data can promptly adapt to changes in customer interests and provide more accurate recommendations.

[0052] Furthermore, after the step of obtaining the second product set of the target customer, the method may further include: processing the historical purchase data of each customer through an association rule mining algorithm to obtain product purchase association information; and adding purchase-related products to the second product set based on the product purchase association information.

[0053] Specifically, association rule mining is a technique in data mining that is used to discover relationships between different things. It works by discovering relationships or "rules" between "item sets" from a large amount of data. The most common algorithms are the Apriori algorithm and the FP-growth algorithm, which can mine rules such as "people who buy A also usually buy B." This algorithm can be applied to scenarios such as retail and recommendation systems to discover patterns and relationships hidden in data.

[0054] By analyzing the historical purchase data of all customers and using the association rule mining algorithm, we can discover the potential purchase relationship between products and obtain product purchase association information. For example, if a large number of customers also buy "mobile phones" when they buy "earphones", then there is a certain purchase association between "earphones" and "mobile phones".

[0055] Product purchase association information can be a rule such as "When purchasing A, B is often purchased as well." Each rule can have a certain support (indicating the frequency of the rule appearing in the purchase data) and confidence (indicating the probability of the rule being true), which helps to evaluate the strength and reliability of the association rule.

[0056] Based on the product purchase association information obtained, the associated products are added to the second product set of the target customer. For example, if the target customer has browsed or purchased a product, the system will recommend some related products based on the association rules of the product in the historical purchase data. For example, if the customer has browsed smart watches, and the purchase history shows that "smart watches" and "sports shoes" are strongly associated, then the sports shoes are added to the second product set, thereby updating the second product set to form a richer product set. This can include not only the current products that the customer is interested in, but also related products that the customer is potentially interested in, further improving the accuracy and diversity of the recommendations.

[0057] In this embodiment, the historical purchase data of each customer is processed through an association rule mining algorithm, and product purchase association information is introduced. This can expand on the existing product recommendations and recommend some potentially related but not browsed or purchased products to target customers, thereby improving the diversity and coverage of the recommendation results and improving the accuracy of the recommendations.

[0058] Furthermore, the above-mentioned step of embedding attributes and knowledge graphs into the product information of each product to obtain product portraits of each product may include: preprocessing the unstructured data in the product information of each product to obtain structured data of each product; extracting attribute information from the structured data of each product to obtain structured attribute data of each product; mapping the structured attribute data of each product into attribute embedding representations; based on the attribute embedding representations of each product, determining the associations between the products through similarity calculations and clustering algorithms, and constructing a product knowledge graph based on the obtained associations; fusing the attribute embedding representations and knowledge graph embedding representations of each product to obtain product portraits of each product.

[0059] Specifically, product information contains unstructured data (e.g., product descriptions, user reviews, pictures, etc.), which are first converted into structured data through preprocessing operations such as text processing and image processing. For example, the text in the product description can be extracted through natural language processing technology to extract keywords, sentiment analysis, and entity recognition, and the picture can be extracted through computer vision algorithms to extract features and converted into structured data.

[0060] The attribute information of the product is extracted from the structured data of each product to obtain the structured attribute data of each product, such as brand, category, price, size, etc.

[0061] Map the structured attribute data of each product into an attribute embedding representation. For example, for information such as the brand, type, and size of a product, they are converted into vector representations through the embedding layer. The purpose of embedding is to transform the discrete attributes of a product into a low-dimensional representation with semantics, capturing the semantic relationship between attributes for subsequent calculations and analysis. This process is usually learned through a neural network to obtain better representation capabilities.

[0062] Using similarity calculation methods, we can calculate the similarity between different products based on the product attribute embedding representation. For example, we can use cosine similarity, Manhattan distance and other metrics to calculate the similarity between product vectors, so as to determine which products have similarities and associations in attributes.

[0063] Clustering algorithms can also be used to group products. For example, through K-means clustering, products with similar characteristics can be grouped into the same category, and the products in each category have an associated relationship.

[0064] Based on the obtained association relationships, a product knowledge graph can be constructed, which contains information such as similarity relationships and association relationships between products. A knowledge graph is a knowledge base represented by a graph structure, where nodes represent entities (such as products) and edges represent relationships between entities. In a product recommendation system, a knowledge graph integrates the relationships between various products (such as similarity, purchase association, etc.) to establish an association network between products, helping the recommendation system to better understand the relationships between products.

[0065] After building the knowledge graph, the attribute embedding representation of the product is fused with the knowledge graph embedding representation of the product in the knowledge graph. The fusion method can be splicing, weighted average, etc. Through fusion, a product portrait of each product is obtained, which contains the basic attribute information of the product and combines the relationship and similarity information between it and other products.

[0066] In this embodiment, attribute information is extracted from product information, and the association relationship between products is determined based on the attribute embedding representation of each product. A product knowledge graph is constructed based on the association relationship, revealing the potential semantic relationship and connection between products. The attribute embedding representation and the knowledge graph embedding representation of each product are comprehensively embedded to generate a comprehensive and accurate product portrait. The portrait not only includes the basic attributes of the product, but also combines the similarities and association information between products. This can better understand the complex relationship between products, which is conducive to subsequently improving the relevance and accuracy of product recommendations.

[0067] Furthermore, the above step S204 may include: generating a node sequence on the customer-product interaction graph by a random walk algorithm; inputting the node sequence into a skip-gram model to generate product node features of each product node in the customer-product interaction graph by maximizing the node co-occurrence probability.

[0068] Specifically, a product interaction graph is constructed based on each customer's purchase history data, browsing history and other interaction data. In this graph, the nodes include customer nodes and product nodes. Edges represent customer interactions with products (such as purchase or browsing). Edges between products reflect their degree of association in customer interactions.

[0069] A random walk algorithm is used to walk in the customer-product interaction graph to generate a node sequence. The random walk algorithm is a graph algorithm in which each node selects an adjacent node as the next node to visit according to a certain probability. This method simulates the process of "roaming" in a graph and is widely used in fields such as graph embedding and social network analysis. In a recommendation system, random walks can be used to generate a node sequence of a product interaction graph, thereby revealing potential associations between products. In this application, the algorithm starts from a product node and randomly selects an adjacent node (i.e., a related product) as the next node to visit. By generating multiple node sequences through multiple iterations, the model can capture the relationships and similarities between products.

[0070] A node sequence refers to a series of nodes (product nodes in this application) generated by a random walk algorithm. The order of these nodes reflects their relationship and interaction frequency in the graph. Through the node sequence, the potential correlation between different products can be revealed.

[0071] The node sequence is input into the Skip-gram model. The Skip-gram model is a model based on word embedding (Word2Vec) in natural language processing, which is used to learn the vector representation of vocabulary. In the recommendation system, the Skip-gram model is used to learn node (product) features. Skip-gram generates an embedding representation of each node by maximizing the probability of co-occurrence of adjacent nodes. In the graph embedding task, it can capture the semantic relationship between nodes.

[0072] In this application, the goal of the Skip-gram model is to learn the embedding vector (i.e., product node feature) of each product node by maximizing the probability of node co-occurrence. This vector can be used to represent the semantic relationship of the product in the interaction graph. The model continuously adjusts the embedding vector so that nodes that frequently appear together in the sequence have similar vector representations, reflecting their similarity in the interaction graph.

[0073] The generated product node features not only reflect the static information of the product (such as category, price, etc.), but also include dynamic association information learned through customer behavior and interaction relationships.

[0074] In this embodiment, by using the random walk algorithm and the Skip-gram model, the feature representation of the product node can more accurately capture the correlation between products. This feature learned through customer behavior and interaction data can more accurately reflect the potential similarities of products, thereby enhancing the accuracy of the recommendation system's understanding and prediction of customer preferences.

[0075] Furthermore, the above-mentioned intelligent product selection management method based on machine learning may also include: for each product node in the customer product interaction diagram, determining the neighbor nodes of the product node; calculating the node similarity between the product node and each neighbor node; determining the attention weight of each neighbor node based on the obtained node similarity; aggregating the product node features of each neighbor node based on the obtained attention weight to obtain aggregated neighbor information; and updating the product node features of the product node based on the aggregated neighbor information.

[0076] Specifically, for each product node in the customer-product interaction graph, the nodes directly connected to it are found in the customer-product interaction graph. These nodes are the neighbor nodes of the target node. Neighbor nodes usually refer to those products that have a strong connection with the target product in the customer's purchase, browsing or other interactive behaviors. For example, a product A may be a neighbor of products B, C, D, etc., because they are often purchased or browsed by the same group of customers.

[0077] For each pair of product nodes and their neighbor nodes, similarity is calculated, for example, by cosine similarity or Jaccard similarity based on customer interactions. The higher the similarity, the closer the relationship between the two products in customer interactions.

[0078] According to the calculated node similarity, an attention weight is assigned to each neighbor node. The attention weight reflects the importance of the neighbor node in the feature update of the target product node. Neighbor nodes with high similarity will have higher weights, and neighbor nodes with low similarity will have lower weights.

[0079] The attention weights are used to weight the features of neighbor nodes to obtain a new feature representation, namely the aggregated neighbor information. This weighted feature can better capture the influence of neighbor nodes in the recommendation, thereby providing more accurate product recommendations.

[0080] The product node features of the product node are updated according to the aggregated neighbor information. After the update, the product node features not only reflect its own information, but also integrate the information of its related neighbor nodes, which can help the system to more accurately understand and recommend products that users may be interested in.

[0081] In this embodiment, the neighbor nodes of the product node are determined, the node similarity between the product node and each neighbor node is calculated, and the attention weight of each neighbor node is determined according to the node similarity. By introducing the features of the neighbor nodes and weightedly aggregating these features according to the node similarity and the attention mechanism, it is beneficial to deeply explore the potential relationship between products. The product node feature representation of the product is more comprehensive, which can more accurately reflect the user's potential interests and capture the deep relationship between products, which is conducive to providing more personalized and accurate product recommendations.

[0082] Furthermore, the above step S205 may include: for each product, extracting a product embedding vector of the product from the product portrait of the product; merging the product embedding vector with a product node feature of a product node corresponding to the product to obtain a comprehensive product feature of the product.

[0083] Specifically, for each product, the product embedding vector is extracted from the previously generated product portrait. It is a low-dimensional vector representation of various product information generated through embedding technology. It captures various characteristics of the product, such as the static characteristics of the product (such as brand, category, price), and may also integrate user behavior data (such as purchase history, browsing history).

[0084] Then obtain the product node features of the product node corresponding to the product, and merge the product embedding vector with the product node features. This combination is to achieve comprehensive information integration. Product node features can include basic information and historical behavior data of the product. The product embedding vector captures the potential relationship between the product and other products and users in a more complex way. When merging, the two types of information can be combined by splicing, weighting, etc. to form a more multi-dimensional comprehensive feature representation, that is, the comprehensive product features of the product.

[0085] The comprehensive product features are the final representation of the product used in the recommendation system. They can describe each product more accurately. Combining static attributes and dynamically generated behavioral features, they can provide richer information for subsequent recommendation models and help the system make more accurate recommendations to users.

[0086] In this embodiment, the product embedding vector of the product is combined with the product node features, and the generated product comprehensive features can more comprehensively reflect the multi-dimensional information of the product. This fused feature enables the recommendation system to more accurately understand the potential relationship between the product and the user, and can simultaneously capture the explicit attributes and implicit relationships of the product, thereby improving the depth of understanding of the product, providing more accurate personalized recommendations, and enhancing the accuracy of recommendations and user satisfaction.

[0087] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0088] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0089] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0090] Further references Figure 3 , as a response to the above Figure 2The present application provides an embodiment of an intelligent product selection management device based on machine learning, and the device embodiment is similar to Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0091] like Figure 3 As shown, the intelligent product selection management device 300 based on machine learning described in this embodiment includes: a preference generation module 301, a product generation module 302, an interaction construction module 303, a feature generation module 304, a comprehensive generation module 305, a combination calculation module 306 and a product recommendation module 307, wherein: The preference generation module 301 is used to generate a customer preference profile of each customer based on the historical purchase data and browsing data of each customer through a preset profile generation algorithm.

[0092] The product generation module 302 is used to obtain product information of each product, and embed attributes and knowledge graphs into the product information of each product to obtain a product portrait of each product.

[0093] The interaction construction module 303 is used to construct a customer-product interaction graph based on the obtained historical purchase data and browsing data. The customer-product interaction graph includes customer nodes and product nodes.

[0094] The feature generation module 304 is used to perform representation learning on the customer-product interaction graph to obtain product node features of each product node.

[0095] The comprehensive generation module 305 is used to generate comprehensive product features of each product based on the product portrait of each product and the product node features of its corresponding product node.

[0096] The combination calculation module 306 is used to combine the customer preference profile of the target customer and the comprehensive product features of each product to obtain a combination feature, and input each combination feature into a matching model to obtain the interest degree between the target customer and each product.

[0097] The product recommendation module 307 is used to determine a target product from among the products according to the obtained interest level, and to make product recommendations to the target customers according to the determined target product.

[0098] In this embodiment, a customer preference portrait of each customer is generated through historical purchase data and browsing data, and the product information of each product is embedded in attributes and knowledge graphs to generate high-quality product portraits; a customer-product interaction graph is constructed and representation learning is performed to further enhance the implicit correlation representation between product nodes; the product portrait of each product and the product node features of its corresponding product nodes are combined to convert the complex relationship between customers and products into comprehensive product features represented in multiple dimensions; the customer preference portrait and the comprehensive product features are input into a matching model to accurately calculate the interest level, thereby achieving personalized recommendations, improving the accuracy of product selection and recommendations, and improving the conversion rate of product recommendations and customer satisfaction.

[0099] In some optional implementations of this embodiment, the preference generation module 301 may include: a customer determination submodule, a purchase determination submodule, a similarity calculation submodule, a browsing determination submodule, and a preference generation submodule, wherein: The customer determination submodule is used to calculate the customer similarity between each customer through a collaborative filtering algorithm based on the historical purchase data of each customer, and determine similar customers of the target customer among each customer according to the obtained customer similarity, and the target customer comes from each customer.

[0100] The purchase determination submodule is used to determine the purchase preference products of each similar customer based on the historical purchase data of each similar customer, and to construct the first product set of the target customer based on each purchase preference product.

[0101] The similarity calculation submodule is used to calculate the product similarity between each product through a content-based recommendation algorithm based on the browsing data of each customer.

[0102] The browsing determination submodule is used to determine the browsing preference products similar to the current browsing products of the target customer according to the obtained product similarities, and add each browsing preference product to the first product set to obtain the second product set of the target customer.

[0103] The preference generation submodule is used to generate a customer preference profile of the target customer based on the target customer's second product set, historical purchase data and browsing data, and obtain the customer preference profile of each customer.

[0104] In this embodiment, by combining historical purchase data and browsing data, recommendations can be generated based on the customer's past behavior, and the recommended content can be adjusted according to the current browsing preferences, thereby expanding the diversity of recommendations and facilitating more personalized and accurate product recommendations; real-time browsing data can promptly adapt to changes in customer interests and provide more accurate recommendations.

[0105] In some optional implementations of this embodiment, the preference generating module 301 may further include: an association determining submodule and an association adding submodule, wherein: The association determination submodule is used to process the historical purchase data of each customer through an association rule mining algorithm to obtain product purchase association information.

[0106] The association adding submodule is used to add purchase-related products to the second product set according to the product purchase association information.

[0107] In this embodiment, the historical purchase data of each customer is processed through an association rule mining algorithm, and product purchase association information is introduced. This can expand on the existing product recommendations and recommend some potentially related but not browsed or purchased products to target customers, thereby improving the diversity and coverage of the recommendation results and improving the accuracy of the recommendations.

[0108] In some optional implementations of this embodiment, the product generation module 302 may include: a preprocessing submodule, an attribute extraction submodule, an attribute mapping submodule, a relationship determination submodule, and an embedding fusion submodule, wherein: The preprocessing submodule is used to preprocess the unstructured data in the product information of each product to obtain the structured data of each product.

[0109] The attribute extraction submodule is used to extract attribute information from the structured data of each product to obtain the structured attribute data of each product.

[0110] The attribute mapping submodule is used to map the structured attribute data of each product into attribute embedding representation.

[0111] The relationship determination submodule is used to determine the association relationship between products based on the attribute embedding representation of each product through similarity calculation and clustering algorithm, and construct a product knowledge graph based on the obtained association relationship.

[0112] The embedding fusion submodule is used to fuse the attribute embedding representation of each product with the knowledge graph embedding representation to obtain the product portrait of each product.

[0113] In this embodiment, attribute information is extracted from product information, and the association relationship between products is determined based on the attribute embedding representation of each product. A product knowledge graph is constructed based on the association relationship, revealing the potential semantic relationship and connection between products. The attribute embedding representation and the knowledge graph embedding representation of each product are comprehensively embedded to generate a comprehensive and accurate product portrait. The portrait not only includes the basic attributes of the product, but also combines the similarities and association information between products. This can better understand the complex relationship between products, which is conducive to subsequently improving the relevance and accuracy of product recommendations.

[0114] In some optional implementations of this embodiment, the feature generation module 304 may include: a sequence generation submodule and a feature determination submodule, wherein: The sequence generation submodule is used to generate node sequences on the customer-product interaction graph through a random walk algorithm.

[0115] The feature determination submodule is used to input the node sequence into the skip-gram model to generate product node features of each product node in the customer-product interaction graph by maximizing the node co-occurrence probability.

[0116] In this embodiment, by using the random walk algorithm and the Skip-gram model, the feature representation of the product node can more accurately capture the correlation between products. This feature learned through customer behavior and interaction data can more accurately reflect the potential similarities of products, thereby enhancing the accuracy of the recommendation system's understanding and prediction of customer preferences.

[0117] In some optional implementations of this embodiment, the feature generation module 304 may further include: a neighbor determination submodule, a similarity calculation submodule, a weight determination submodule, an aggregation determination submodule, and a feature update submodule, wherein: The neighbor determination submodule is used to determine the neighbor nodes of each product node in the customer-product interaction graph.

[0118] The similarity calculation submodule is used to calculate the node similarity between the product node and each neighbor node.

[0119] The weight determination submodule is used to determine the attention weight of each neighbor node according to the obtained node similarity.

[0120] The aggregation determination submodule is used to aggregate the product node features of each neighbor node according to the obtained attention weight to obtain aggregated neighbor information.

[0121] The feature update submodule is used to update the product node features of the product node according to the aggregated neighbor information.

[0122] In this embodiment, the neighbor nodes of the product node are determined, the node similarity between the product node and each neighbor node is calculated, and the attention weight of each neighbor node is determined according to the node similarity. By introducing the features of the neighbor nodes and weightedly aggregating these features according to the node similarity and the attention mechanism, it is beneficial to deeply explore the potential relationship between products. The product node feature representation of the product is more comprehensive, which can more accurately reflect the user's potential interests and capture the deep relationship between products, which is conducive to providing more personalized and accurate product recommendations.

[0123] In some optional implementations of this embodiment, the comprehensive generation module 305 may include: a vector extraction submodule and a merging submodule, wherein: The vector extraction submodule is used to extract the product embedding vector of each product from the product portrait of the product.

[0124] The merging submodule is used to merge the product embedding vector with the product node feature of the product node corresponding to the product to obtain the comprehensive product feature of the product.

[0125] In this embodiment, the product embedding vector of the product is combined with the product node features, and the generated product comprehensive features can more comprehensively reflect the multi-dimensional information of the product. This fused feature enables the recommendation system to more accurately understand the potential relationship between the product and the user, and can simultaneously capture the explicit attributes and implicit relationships of the product, thereby improving the depth of understanding of the product, providing more accurate personalized recommendations, and enhancing the accuracy of recommendations and user satisfaction.

[0126] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0127] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected and communicated through a system bus. It should be noted that the figure only shows a computer device 4 with a memory 41, a processor 42, and a network interface 43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.

[0128] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0129] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped on the computer device 4. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as a computer-readable instruction of an intelligent product selection management method based on machine learning, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0130] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run computer-readable instructions stored in the memory 41 or process data, such as computer-readable instructions for running the intelligent product selection management method based on machine learning.

[0131] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0132] The computer device provided in this embodiment can execute the above-mentioned intelligent product selection management method based on machine learning. Here, the intelligent product selection management method based on machine learning can be an intelligent product selection management method based on machine learning in each of the above-mentioned embodiments.

[0133] In this embodiment, a customer preference portrait of each customer is generated through historical purchase data and browsing data, and the product information of each product is embedded in attributes and knowledge graphs to generate high-quality product portraits; a customer-product interaction graph is constructed and representation learning is performed to further enhance the implicit correlation representation between product nodes; the product portrait of each product and the product node features of its corresponding product nodes are combined to convert the complex relationship between customers and products into comprehensive product features represented in multiple dimensions; the customer preference portrait and the comprehensive product features are input into a matching model to accurately calculate the interest level, thereby achieving personalized recommendations, improving the accuracy of product selection and recommendations, and improving the conversion rate of product recommendations and customer satisfaction.

[0134] The present application also provides another embodiment, namely, providing a computer-readable storage medium, which stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of an intelligent product selection management method based on machine learning as described above.

[0135] In this embodiment, a customer preference portrait of each customer is generated through historical purchase data and browsing data, and the product information of each product is embedded in attributes and knowledge graphs to generate high-quality product portraits; a customer-product interaction graph is constructed and representation learning is performed to further enhance the implicit correlation representation between product nodes; the product portrait of each product and the product node features of its corresponding product nodes are combined to convert the complex relationship between customers and products into comprehensive product features represented in multiple dimensions; the customer preference portrait and the comprehensive product features are input into a matching model to accurately calculate the interest level, thereby achieving personalized recommendations, improving the accuracy of product selection and recommendations, and improving the conversion rate of product recommendations and customer satisfaction.

[0136] The present application provides an intelligent product selection management system based on machine learning, which is also a recommendation system, including a computer device and a computer-readable storage medium, wherein the computer device has an intelligent product selection management device based on machine learning.

[0137] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic 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, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0138] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. An intelligent product selection management method based on machine learning, characterized in that: The steps include: Based on the historical purchase data and browsing data of each customer, a customer preference profile of each customer is generated by a preset profile generation algorithm; Obtain product information of each product, and embed attributes and knowledge graphs of the product information of each product to obtain product portraits of each product; Constructing a customer-product interaction graph based on the obtained historical purchase data and browsing data, wherein the customer-product interaction graph includes customer nodes and product nodes; Performing representation learning on the customer-product interaction graph to obtain product node features of each product node; Generate comprehensive product features of each product according to the product profile of each product and the product node features of its corresponding product node; Combining the customer preference profile of the target customer and the comprehensive product features of each product to obtain combined features, and inputting each combined feature into a matching model to obtain the interest degree between the target customer and each product; A target product is determined from among the products according to the obtained interest level, and a product recommendation is made to the target customer according to the determined target product.

2. According to claim 1, a method for intelligent product selection management based on machine learning is characterized in that: The step of generating the customer preference portrait of each customer by a preset portrait generation algorithm based on the historical purchase data and browsing data of each customer includes: Based on the historical purchase data of each customer, the customer similarity between the customers is calculated by a collaborative filtering algorithm, and similar customers of the target customer are determined from among the customers according to the obtained customer similarity, wherein the target customer comes from the customers; Determine the purchase preference products of each similar customer according to the historical purchase data of each similar customer, and construct the first product set of the target customer according to each purchase preference product; Based on the browsing data of each customer, calculating the product similarity between each product by a content-based recommendation algorithm; Determine browsing preference products similar to the current browsing product of the target customer according to the obtained product similarity, and add each browsing preference product to the first product set to obtain a second product set of the target customer; A customer preference profile of the target customer is generated based on the second product set, historical purchase data and browsing data of the target customer, and a customer preference profile of each customer is obtained.

3. The intelligent product selection management method based on machine learning according to claim 2 is characterized in that: After the step of obtaining the second product set of the target customer, the method further includes: Process the historical purchase data of each customer through an association rule mining algorithm to obtain product purchase association information; According to the product purchase association information, purchase-associated products are added to the second product set.

4. The intelligent product selection management method based on machine learning according to claim 1 is characterized in that: The step of embedding attributes and knowledge graphs into the product information of each product to obtain a product portrait of each product includes: Preprocessing the unstructured data in the product information of each product to obtain the structured data of each product; Extracting attribute information from the structured data of each product to obtain structured attribute data of each product; Mapping the structured attribute data of each product into attribute embedding representations respectively; Based on the attribute embedding representation of each product, determine the association relationship between the products through similarity calculation and clustering algorithm, and construct a product knowledge graph according to the obtained association relationship; The attribute embedding representation of each product and the knowledge graph embedding representation are fused to obtain the product portrait of each product.

5. The intelligent product selection management method based on machine learning according to claim 1 is characterized in that: The step of performing representation learning on the customer-product interaction graph to obtain product node features of each product node comprises: Generate a node sequence on the customer-product interaction graph by a random walk algorithm; The node sequence is input into a skip-gram model to generate product node features of each product node in the customer-product interaction graph by maximizing the node co-occurrence probability.

6. The intelligent product selection management method based on machine learning according to claim 5 is characterized in that: The method further comprises: For each product node in the customer-product interaction graph, determining a neighbor node of the product node; Calculating the node similarity between the product node and each neighbor node; Determine the attention weight of each neighbor node according to the obtained node similarity; Aggregating the product node features of each neighbor node according to the obtained attention weights to obtain aggregated neighbor information; The product node feature of the product node is updated according to the aggregated neighbor information.

7. The intelligent product selection management method based on machine learning according to claim 1 is characterized in that: The step of generating the comprehensive product features of each product according to the product portrait of each product and the product node features of its corresponding product node comprises: For each product, extract a product embedding vector of the product from the product portrait of the product; The product embedding vector is combined with the product node feature of the product node corresponding to the product to obtain the comprehensive product feature of the product.

8. An intelligent product selection management device based on machine learning, characterized in that: include: A preference generation module, used to generate a customer preference profile of each customer through a preset profile generation algorithm based on the historical purchase data and browsing data of each customer; A product generation module is used to obtain product information of each product, and embed attributes and knowledge graphs into the product information of each product to obtain a product profile of each product; An interaction construction module, used to construct a customer-product interaction graph based on the obtained historical purchase data and browsing data, wherein the customer-product interaction graph includes customer nodes and product nodes; A feature generation module, used for performing representation learning on the customer-product interaction graph to obtain product node features of each product node; A comprehensive generation module, used to generate comprehensive product features of each product according to the product profile of each product and the product node features of its corresponding product node; A combination calculation module, used to combine the customer preference profile of the target customer and the comprehensive product features of each product to obtain a combination feature, and input each combination feature into a matching model to obtain the interest degree between the target customer and each product; The product recommendation module is used to determine a target product from among the products according to the obtained interest level, and to make product recommendations to the target customer according to the determined target product.

9. A computer device, characterized in that: It includes a memory and a processor, the memory stores computer-readable instructions, and the processor implements the steps of an intelligent product selection management method based on machine learning as described in any one of claims 1 to 7 when executing the computer-readable instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of an intelligent product selection management method based on machine learning as described in any one of claims 1 to 7.

Images