Product recommendation method and device, computer equipment and storage medium

Abstract

The embodiment of the invention belongs to the field of artificial intelligence, and relates to a product recommendation method, which comprises the following steps: obtaining a candidate product library, and obtaining user data of all users and product data of all candidate products; determining a candidate product corresponding to each user in a candidate product library according to a preset recall strategy; constructing an interaction matrix and a weight matrix according to the user data, calculating a user feature vector of each user according to the interaction matrix, the weight matrix and the user data, and calculating a product feature vector of each candidate product according to the interaction matrix, the weight matrix and the product data; calculating a matching degree between each user and each candidate product according to the obtained weight matrix, the user feature vector and the product feature vector, and determining a target product of each user based on the obtained matching degree; and according to the target product of each user, carrying out product recommendation on each user. The accuracy of product recommendation is improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a product recommendation method, device, computer equipment and storage medium. Background Art

[0002] With the development of Internet technology and the popularization of e-commerce, recommendation systems are increasingly widely used on major platforms. Recommendation systems aim to provide users with personalized product recommendations, such as various commodities, services or content, thereby improving users' purchasing experience and the operating efficiency of the platform.

[0003] However, existing recommendation systems still have some deficiencies and defects in practical applications. Many recommendation systems have problems in displaying product information flows, and the displayed content is often not what users really need or are interested in, resulting in a low click-through rate of the information flow. In addition, when recommending products, existing recommendation systems often cannot accurately identify and recommend products that users may be interested in, and the relevant recommended content cannot meet their needs. In the fields of finance and insurance, the application of recommendation systems also faces similar challenges. There are many types of financial and insurance products and user needs are diverse, which further reduces the accuracy of product recommendations. Summary of the invention

[0004] The purpose of the embodiments of the present application is to provide a product recommendation method, apparatus, computer device and storage medium to solve the problem of low accuracy of product recommendations.

[0005] In order to solve the above technical problems, the present application embodiment provides a product recommendation method, which adopts the following technical solution:

[0006] Obtain a candidate product library;

[0007] According to a preset recall strategy, determining a candidate product corresponding to each user in the candidate product library;

[0008] Obtain user data of all users and product data of all candidate products;

[0009] constructing an interaction matrix and a weight matrix according to the user data, and calculating a user feature vector of each user according to the interaction matrix, the weight matrix and the user data, and calculating a product feature vector of each candidate product according to the interaction matrix, the weight matrix and the product data;

[0010] Calculate the matching degree between each user and each candidate product according to the obtained weight matrix, user feature vector and product feature vector, and determine the target product of each user based on the obtained matching degree;

[0011] According to the target product of each user, product recommendations are made to each user.

[0012] In order to solve the above technical problems, the embodiment of the present application further provides a product recommendation device, which adopts the following technical solution:

[0013] A product library acquisition module is used to acquire a candidate product library;

[0014] A candidate determination module, used to determine the candidate product corresponding to each user in the candidate product library according to a preset recall strategy;

[0015] A data acquisition module is used to acquire user data of all users and product data of all candidate products;

[0016] a vector calculation module, configured to construct an interaction matrix and a weight matrix according to the user data, and calculate a user feature vector of each user according to the interaction matrix, the weight matrix and the user data, and calculate a product feature vector of each candidate product according to the interaction matrix, the weight matrix and the product data;

[0017] A product determination module, configured to calculate the matching degree between each user and each candidate product according to the obtained weight matrix, user feature vector and product feature vector, and determine the target product of each user based on the obtained matching degree;

[0018] The product recommendation module is used to recommend products to each user according to the target products of each user.

[0019] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0020] Obtain a candidate product library;

[0021] According to a preset recall strategy, determining a candidate product corresponding to each user in the candidate product library;

[0022] Obtain user data of all users and product data of all candidate products;

[0023] constructing an interaction matrix and a weight matrix according to the user data, and calculating a user feature vector of each user according to the interaction matrix, the weight matrix and the user data, and calculating a product feature vector of each candidate product according to the interaction matrix, the weight matrix and the product data;

[0024] Calculate the matching degree between each user and each candidate product according to the obtained weight matrix, user feature vector and product feature vector, and determine the target product of each user based on the obtained matching degree;

[0025] According to the target product of each user, product recommendations are made to each user.

[0026] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0027] Obtain a candidate product library;

[0028] According to a preset recall strategy, determining a candidate product corresponding to each user in the candidate product library;

[0029] Obtain user data of all users and product data of all candidate products;

[0030] constructing an interaction matrix and a weight matrix according to the user data, and calculating a user feature vector of each user according to the interaction matrix, the weight matrix and the user data, and calculating a product feature vector of each candidate product according to the interaction matrix, the weight matrix and the product data;

[0031] Calculate the matching degree between each user and each candidate product according to the obtained weight matrix, user feature vector and product feature vector, and determine the target product of each user based on the obtained matching degree;

[0032] According to the target product of each user, product recommendations are made to each user.

[0033] Compared with the prior art, the embodiments of the present application mainly have the following beneficial effects: first, by combining a variety of recall strategies, candidate products related to the user are screened out from different angles, thereby increasing the diversity and relevance of recommended products; an interaction matrix and a weight matrix are constructed based on user data, the element values ​​of the elements in the interaction matrix represent the degree of preference of each user for each candidate product, and the element values ​​of the elements in the weight matrix represent the weights of the corresponding elements in the interaction matrix; by applying a weighted matrix decomposition algorithm to the interaction matrix and the weight matrix, and then using user data and product data, user feature vectors and product feature vectors are calculated, which effectively captures the implicit features of users and products and improves the accuracy of recommendations; based on the obtained weight matrix, user feature vector and product feature vector, the matching degree between each user and each candidate product is calculated, and the target product of each user is determined based on the obtained matching degree, and the matching degree calculation makes the recommended products more in line with user needs; finally, personalized product recommendations are made based on the target product of each user, which is conducive to improving user satisfaction and click-through rate, and improving the accuracy of product recommendations. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] 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.

[0035] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0036] Figure 2 is a flow chart of an embodiment of a product recommendation method according to the present application;

[0037] Figure 3 is a structural schematic diagram of an embodiment of a product recommendation device according to the present application;

[0038] Figure 4 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0039] 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.

[0040] 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.

[0041] 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.

[0042] like Figure 1As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables.

[0043] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0044] The terminal devices 101, 102, and 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, desktop computers, etc.

[0045] The server 105 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal devices 101 , 102 , and 103 .

[0046] It should be noted that the product recommendation method provided in the embodiment of the present application is generally executed by a server, and accordingly, the product recommendation device is generally set in the server.

[0047] 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 implementation requirements.

[0048] Continue to refer Figure 2 , shows a flow chart of an embodiment of a product recommendation method according to the present application. The product recommendation method comprises the following steps:

[0049] Step S201, obtaining a candidate product library, and obtaining user data of all users and product data of all candidate products.

[0050] In this embodiment, the electronic device (eg, Figure 1The 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, Wi-Fi connection, Bluetooth connection, Wi MAX connection, Zigbee connection, UWB (Ultra Wi-Fi Band) connection, and other wireless connection methods currently known or to be developed in the future.

[0051] Specifically, firstly, a candidate product library is obtained from the database, which includes all products that may be recommended to the user at present. At the same time, the system also needs to obtain the user data of all users (including user basic data, user behavior data, user relationship data, etc.) and the product data of all candidate products.

[0052] Among them, user basic data describes the basic attributes of users. For example, in the field of financial insurance, user basic data includes the user's insurance grouping (for example, it can include purchased auto insurance, purchased non-auto insurance, purchased auto insurance but not purchased non-auto insurance, not purchased auto insurance but purchased non-auto insurance, and no insurance purchased), user car category grouping (for example, it can include new energy car owners, oil car owners, hybrid car owners), user family attributes (married, unmarried, married with children, etc.), user driving experience (0-1 year new car owner, 1-3 years driving experience car owner, 3 years and above car owner, etc.), user car model attributes (sedan, sports car, SUV, family car, etc.), user's membership level in the financial insurance company (not a member, gold, platinum, diamond, supreme, etc.), user gender (male, female), user income (less than 100,000, 100,000-300,000, 300,000-500,000, 500,000 and above), and user location.

[0053] User behavior data is data related to user behavior, which can be classified into explicit data and implicit data. Among them, explicit data can be user ratings and comments on products, etc. Implicit data includes user browsing behavior (which can record the products browsed by users, browsing time, browsing duration, etc.), user interaction behavior (such as sharing products, adding products to cart), user purchasing behavior (which can record the product type, product price, product quantity, etc. of the products purchased by users), user shopping tendencies (such as price-sensitive, brand-sensitive, service attitude-sensitive, positioning-sensitive, etc.), and user search behavior (user search keywords, etc.).

[0054] User relationship data refers to data related to user social interaction and social behavior, including strong relationship data (such as mutual attention between users) and weak relationship data (such as mutual likes and comments between users).

[0055] Product data describes the basic attributes of a product, including product price (regular price, member price, discounted price, price after coupon, etc.), product classification (with 1st, 2nd, and 3rd level categories), and product tags (manually added product tags and data statistical rule tags, both of which are used to describe certain attributes of a product).

[0056] Step S202: determining the candidate product corresponding to each user in the candidate product library according to a preset recall strategy.

[0057] Specifically, the present application pre-sets a variety of recall strategies (such as recalling products based on product popularity, recalling products based on marketing activities, etc.) According to the multi-dimensional recall strategy, multiple candidate products that users may be interested in are screened out from the candidate product library.

[0058] Step S203, constructing an interaction matrix and a weight matrix according to the user data, and calculating a user feature vector of each user according to the interaction matrix, the weight matrix and the user data, and calculating a product feature vector of each candidate product according to the interaction matrix, the weight matrix and the product data.

[0059] Specifically, the interaction matrix R and the weight matrix W are constructed based on the user data. The element value in the interaction matrix represents the preference of each user for each candidate product. Assuming there are two users and two products, a 2×2 interaction matrix is ​​constructed. The weight matrix is ​​the same size as the interaction matrix. The element value of each element in the weight matrix represents the weight of the element at the corresponding position in the interaction matrix. For example, the element r in the interaction matrix R is ij represents the preference value of user i for product j, element r ij The corresponding element in the weight matrix W is w ij , represents the weight of the interaction between user i and product j.

[0060] Based on the weighted matrix decomposition algorithm, the user feature vector of each user can be calculated according to the interaction matrix, weight matrix and user data. The user feature vector is a vector representation of the user data and contains the implicit features of the user.

[0061] Based on the weighted matrix decomposition algorithm, the product feature vector of each candidate product can also be calculated according to the interaction matrix, weight matrix and product data. The product feature vector is a vector representation of the product data and contains the implicit features of the product.

[0062] Step S204, calculating the matching degree between each user and each candidate product according to the obtained weight matrix, user feature vector and product feature vector, and determining the target product of each user based on the obtained matching degree.

[0063] Specifically, the matching degree between each user and each candidate product can be calculated using the user feature vector, product feature vector and weight matrix calculated above. As mentioned above, the element w in the weight matrix W is ij represents the weight of the interaction between user i and product j. When calculating the matching degree between user i and product j, the corresponding element w is obtained from the weight matrix. ij , according to the element w ij , the user feature vector of user i and the product feature vector of product j are calculated.

[0064] The matching degree reflects the user's interest in the candidate product and the degree of suitability. The higher the matching degree, the greater the user's interest in the product, or the higher the matching degree between the user and the product. Based on the calculated matching degree, a set of target products is determined for each user, and these products will be used as candidates for the final recommendation to the user.

[0065] It should be noted that here, for each user, the matching degree between the user and the candidate products of all users is calculated respectively. Different users have their own set of matching degree data and target products.

[0066] Step S205: recommend products to each user based on the target products of each user.

[0067] Specifically, based on the target products of each user determined in the previous step, the product recommendation system recommends these target products to the corresponding users. During the recommendation process, the product recommendation system can also combine business rules (such as product launch status, inventory status, etc.) to further optimize the recommendation results to ensure that the recommended products not only meet the user's interests, but also meet actual business needs.

[0068] In this embodiment, firstly, by combining multiple recall strategies, candidate products related to the user are screened out from different angles, thereby increasing the diversity and relevance of the recommended products; an interaction matrix and a weight matrix are constructed based on user data, the element values ​​of the elements in the interaction matrix represent the preference of each user for each candidate product, and the element values ​​of the elements in the weight matrix represent the weights of the corresponding elements in the interaction matrix; by applying a weighted matrix decomposition algorithm to the interaction matrix and the weight matrix, and then using user data and product data, user feature vectors and product feature vectors are calculated, which effectively captures the implicit features of users and products and improves the accuracy of recommendations; based on the obtained weight matrix, user feature vector and product feature vector, the matching degree between each user and each candidate product is calculated, and the target product of each user is determined based on the obtained matching degree, and the matching degree calculation makes the recommended products more in line with user needs; finally, personalized product recommendations are made based on the target product of each user, which is conducive to improving user satisfaction and click-through rate, and improving the accuracy of product recommendations.

[0069] Furthermore, the above step S202 may include: determining the first recalled product in the candidate product library based on product popularity data; determining the second recalled product in the candidate product library based on product activity data; performing product-based collaborative filtering on each user to determine the third recalled product of each user in the candidate product library; performing user-based collaborative filtering on each user to determine the fourth recalled product of each user in the candidate product library; and performing recall adjustments on the recalled products of each user respectively to obtain candidate products corresponding to each user.

[0070] Specifically, the first recalled product is determined in the candidate product library based on the product popularity data. Product popularity data may include the number of views, clicks, purchases, etc. of the product. Based on this data, the most popular product can be obtained and the first recalled product can be obtained. This strategy ensures that the recalled product has a high degree of popularity and attention.

[0071] Based on the product activity data, the second recalled product is determined in the candidate product library. The product activity data may include promotional information, discount activities, limited-time special sales, etc. Based on the product activity data, the products currently participating in the promotional activities can be determined to obtain the second recalled product. This strategy ensures that the recalled product has high appeal and purchase incentives.

[0072] Product popularity data and product activity data can be obtained separately, such as from a big data platform, or determined based on product tags in product data. Product tags include manually added product tags, as well as data statistical rule tags. For example, if the merchant corresponding to the product is related to helping farmers, you can add an agricultural assistance tag to the product; for new products, you can add a new tag; you can also add a good review tag based on the good reviews, add a preferred tag based on product assessment data, and add a gold list tag to products that are configured with a list and ranked high; add corresponding tags according to the product's views, sales, sales within a preset time period, etc.

[0073] Product-based collaborative filtering (I temCF) is performed on each user to determine the third recalled product for each user in the candidate product library. Product-based collaborative filtering makes recommendations based on product similarity. If a user shows interest in certain products (browsing, purchasing, adding to cart, etc.), the system will find other products similar to these products for recall. This strategy ensures that the recalled products are highly correlated with the user's historical preferences.

[0074] User-based collaborative filtering (UserCF) is performed on each user to determine the fourth recalled product for each user in the candidate product library. User-based collaborative filtering recommends based on user similarities and recalls products favored by other users who have similar interests to the current user. This strategy ensures that the recalled products can meet the user's potential interests and needs.

[0075] Recall adjustments are performed on each user's recalled products to obtain candidate products corresponding to each user. Recall adjustments will be combined with business rules, user data, product data, etc. to optimize each user's recalled products, such as deduplication, sorting adjustment, and screening, to ensure that the candidate product library meets both user interests and business needs.

[0076] In this embodiment, product popularity data and product activity data are combined, as well as product-based collaborative filtering and user-based collaborative filtering to provide users with richer and more personalized product recommendations, meet users' needs for popular products and promotional products, and meet users' potential interests, thereby ensuring the accuracy of subsequent product recommendations.

[0077] Furthermore, the above-mentioned step of constructing an interaction matrix and a weight matrix based on user data may include: extracting user behavior data from the user data; obtaining the interaction factor and interaction weight factor between each user and each candidate product based on the user behavior data; calculating the interaction score between each user and each candidate product based on the obtained interaction factor, and constructing an interaction matrix based on all the interaction scores; calculating the interaction weight between each user and each candidate product based on the obtained interaction weight factor, and constructing a weight matrix based on all the interaction weights.

[0078] Specifically, user behavior data is extracted from the user data, and then the interaction factor and interaction weight factor between each user and each candidate product are obtained from the user behavior data.

[0079] Among them, the interaction factors may include rating data (users' ratings of products), browsing data (number of times or duration of users' browsing of products), purchase data (number of times or amount of money users purchase products), and click data (number of times users click on products). Weights can be set for different types of interaction factors, and based on various types of interaction factors and their weights, weighted scores are calculated as interaction scores between users and candidate products.

[0080] Based on the interaction scores of all users for all candidate products, the interaction matrix R can be constructed. In the interaction matrix, rows represent users, columns represent products, and the element value (i.e., interaction score) of the element in the interaction matrix represents the user's preference for the candidate product. For example, the element r in the interaction matrix R ij represents the preference value of user i for product j.

[0081] The interaction weight factors include time decay factor (as time goes by, the interaction weight between users and products gradually decreases, and recent interaction behaviors are more important than past interaction behaviors), interaction frequency (the higher the interaction frequency between users and products, the greater the weight; for example, if a user clicks or buys a product multiple times, it indicates that the product is more important to the user) and interaction type (different types of interaction behaviors are assigned different weights, for example, purchase behavior has a greater weight than browsing behavior). The interaction weights between users and candidate products can be obtained by weighted summing of various interaction weight factors.

[0082] According to the interaction weights of all users for all candidate products, a weight matrix W is constructed. The weight matrix is ​​the same size as the interaction matrix. The value of each element in the weight matrix represents the weight of the element at the corresponding position in the interaction matrix, that is, the importance of each interaction. For example, the element r in the interaction matrix R ij represents the preference value of user i for product j, element r ij The corresponding element in the weight matrix W is w ij , represents the weight of the interaction between user i and product j.

[0083] In this embodiment, the interaction factors and interaction weight factors in the user behavior data are extracted to more comprehensively capture the interaction information between the user and the product; the interaction score between each user and each candidate product is calculated by weighted summing the interaction factors, reflecting the user's comprehensive interest in the product; the interaction intensity between the user and the product is reflected by weighted summing the interaction weight factors of different interaction behaviors; according to the interaction score and the interaction weight factor, the interaction matrix and the weight matrix are respectively constructed, which provide a data basis for the subsequent feature vector calculation and matching degree calculation.

[0084] Furthermore, the steps of calculating the user feature vector of each user according to the interaction matrix, the weight matrix and the user data, and calculating the product feature vector of each candidate product according to the interaction matrix, the weight matrix and the product data may include: generating an initial user feature vector for each user according to the user data, and generating an initial product feature vector for each candidate product according to the product data; multiplying corresponding elements in the interaction matrix and the weight matrix one by one to obtain a weighted matrix; performing matrix decomposition on the weighted matrix to obtain a user latent vector matrix and a product latent vector matrix; obtaining the user latent vector of each user from the user latent vector matrix, and calculating the user feature vector of each user according to the user latent vector and the initial user feature vector; obtaining the product latent vector of each candidate product from the product latent vector matrix, and calculating the product feature vector of each candidate product according to the product latent vector and the initial product feature vector.

[0085] Specifically, an initial user feature vector of each user is generated based on the user data, and an initial product feature vector of each candidate product is generated based on the product data. For example, the user basic data and product data in the user data are processed by normalization, encoding, TF-IDF or Word2Vec, etc.; the user relationship data in the user data is processed by graph embedding technology.

[0086] Then, the corresponding elements in the interaction matrix R and the weight matrix W are multiplied one by one to obtain the weighted matrix M (M=W*R). This processing method combines the interaction score with the interaction weight to comprehensively reflect the strength of the relationship between the user and the product, or the user's weighted preference for the product.

[0087] Then, weighted matrix decomposition is performed, for example, non-negative matrix decomposition is used to obtain the user latent vector matrix U and the product latent vector matrix V (M≈U·V T ). The user latent vector matrix and the product latent vector matrix represent the feature vectors of users and products in the latent space respectively.

[0088] Each row U of the user latent vector matrix U i Represents a user latent vector of a user, and each column V of the product latent vector matrix V j A product latent vector representing a product. These latent vectors capture the characteristic representation of users and products in the latent space.

[0089] The user feature vector of each user is calculated based on the user latent vector and the initial user feature vector; the product feature vector of each candidate product is calculated based on the product latent vector and the initial product feature vector. The specific calculation can be performed by fusing the (user / product) latent vector and the initial (user / product) feature vector through vector concatenation, weighted fusion, etc.

[0090] In this embodiment, an initial user feature vector of the user is generated based on the user data, and an initial product feature vector of the candidate product is generated based on the product data, so as to capture the explicit features and basic attributes of the user and the product; the corresponding elements in the interaction matrix and the weight matrix are multiplied one by one to obtain a weighted matrix reflecting the user's weighted preference; the user latent vector matrix and the product latent vector matrix are extracted from the weighted matrix through a matrix decomposition algorithm, and the interaction information of the user and the product is compressed into a latent space to capture the potential features in the data; the user latent vector and the initial user feature vector are fused to calculate the user feature vector; the product latent vector and the initial product feature vector are fused to calculate the product feature vector, so as to ensure that the feature vector retains both the explicit features and the latent features, improve the accuracy of the feature vector, and ensure the accuracy of the subsequent matching degree calculation.

[0091] Furthermore, the step of calculating the degree of match between each user and each candidate product based on the obtained weight matrix, user feature vector and product feature vector may include: multiplying the user feature vector and the product feature vector to obtain an initial degree of match between each user and each candidate product; obtaining a weighting factor corresponding to the initial degree of match from the weight matrix; and multiplying the weighting factor and the initial degree of match to obtain a degree of match between each user and each candidate product.

[0092] Specifically, the user feature vector and the product feature vector are multiplied, for example, by a dot product method, to obtain the initial matching degree between each user and each candidate product, reflecting the similarity between user features and product features in multiple dimensions. The higher the initial matching degree value, the greater the user's potential interest in the product.

[0093] Obtain the weighting factor corresponding to the initial matching degree from the weight matrix. i ' and product feature vector V i 'The initial matching degree obtained by multiplication, the weighting factor obtained from the weighting matrix is ​​W ij .

[0094] The weighting factor is multiplied by the corresponding initial match to obtain the final match between each user and each candidate product. The weighting factor can adjust the initial match so that it is more in line with the user's actual preferences. For example, if the user interacts with a product frequently or the type of interaction is important, the weighting factor will amplify the initial match, otherwise it will reduce the initial match.

[0095] In this embodiment, the user feature vector and the product feature vector are multiplied to obtain the initial matching degree, thereby indicating the similarity between the user and the product in multiple dimensions, as well as the user's potential interest in the product; the corresponding weighting factor is then obtained from the weight matrix and multiplied with the initial matching degree, so that the final matching degree more accurately reflects the user's true preference, thereby improving the accuracy and personalization of the matching degree calculation.

[0096] Furthermore, the above-mentioned step of determining the target product of each user based on the obtained matching degree may include: sorting all candidate products in descending order according to the matching degree between each user and all candidate products to obtain a first product queue; making a first adjustment to the first product queue according to the matching degree and product label to obtain a second product queue, wherein the first adjustment includes product screening and sorting adjustment; receiving an adjustment instruction sent by the control terminal, and making a second adjustment to the second product queue according to the adjustment instruction to obtain a third product queue; and determining each product in the third product queue as a target product for each user.

[0097] Specifically, according to the matching degree between each user and all candidate products, all candidate products are sorted in descending order, and products with high matching degree are placed in front, so as to obtain a first product queue.

[0098] According to the matching degree and product label, the first product queue is adjusted to obtain the second product queue. The first adjustment includes product screening and sorting adjustment. Product screening can increase or decrease the products in the first product queue, such as eliminating products with low matching degree, eliminating products that users have purchased within N days, recalling other products of the same type, brand, and store as the current product details page on the user side, recalling products related to festivals / hot spots, recalling seasonal products, and recalling products that users have added to cart but have not purchased within M days. Newly recalled products can have different priorities, and the product sorting can be adjusted according to the priority; for example, seasonal products have the highest priority, and after being recalled, they are placed at the front of the queue. The above adjustments can be made according to product labels, such as recalling products according to product labels, or determining product priorities.

[0099] Receive the adjustment instruction sent by the control terminal. The control terminal can be a terminal held by the operator (such as the insurance product operator of a financial insurance company). The operator can issue adjustment instructions based on the merchant promotion needs, platform strategies, etc., so as to make a second adjustment to the second product queue, such as adding or deleting products, adjusting product sorting, etc., to obtain a third product queue, and determine each product in the third product queue as the user's target product.

[0100] In this embodiment, firstly, according to the matching degree between the user and all the candidate products, all the candidate products are sorted in descending order to obtain a first product queue; a first adjustment is made to the first product queue according to the matching degree and the product label to obtain an optimized second product queue; an adjustment instruction of the control terminal is received and responded to to obtain a third product queue so as to meet the promotion needs or platform strategies; through the optimization and adjustment in multiple dimensions, the relevance and personalization of the recommendation results are improved, so that the target products can better meet the needs of the users and improve the overall accuracy of the recommendations.

[0101] Furthermore, after the above step S205, it may also include: obtaining each user's click information on the target product; determining each user's temporary recommended product based on the click information and each user's target product; and displaying the temporary recommended product on the product page of each user's user terminal.

[0102] Specifically, the user's target product can be displayed through the user terminal held by the user, and the user's click on the displayed target product will generate click information. The user's click information on the target product is obtained, and the temporary recommended product of each user can be determined based on the click information and the target product of each user. The temporary recommended product is a new recommended product generated dynamically, taking into account the user's latest interest changes. In one embodiment, the temporary recommended product can also be generated based on the click information alone.

[0103] On the product page of the user terminal held by the user, temporarily recommended products are displayed. For example, 4 temporarily recommended products are displayed in the form of cards at the bottom of the product page, so that the user can instantly browse and select new recommended products.

[0104] In this embodiment, the user's click information on the target product is obtained to understand the user's latest interests and needs in real time; the user's click information and the target product are combined to dynamically generate temporary recommended products; the temporary recommended products are displayed on the user-side product page, allowing the user to browse and select new recommended products instantly, thereby improving the user's interactive experience and providing a more personalized and timely recommendation service.

[0105] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned user data and product data, the above-mentioned user data and product data can also be stored in a node of a blockchain.

[0106] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0107] 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.

[0108] 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.

[0109] 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).

[0110] 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.

[0111] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of a product recommendation device, 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.

[0112] like Figure 3 As shown, the product recommendation device 300 described in this embodiment includes: an acquisition module 301, a candidate determination module 302, a vector calculation module 303, a product determination module 304 and a product recommendation module 305, wherein:

[0113] The acquisition module 301 is used to acquire a candidate product library, and acquire user data of all users and product data of all candidate products.

[0114] The candidate determination module 302 is used to determine the candidate product corresponding to each user in the candidate product library according to a preset recall strategy.

[0115] The vector calculation module 303 is used to construct an interaction matrix and a weight matrix based on user data, and calculate the user feature vector of each user based on the interaction matrix, the weight matrix and the user data, and calculate the product feature vector of each candidate product based on the interaction matrix, the weight matrix and the product data.

[0116] The product determination module 304 is used to calculate the matching degree between each user and each candidate product according to the obtained weight matrix, user feature vector and product feature vector, and determine the target product of each user based on the obtained matching degree.

[0117] The product recommendation module 305 is used to recommend products to each user based on the target products of each user.

[0118] In this embodiment, firstly, by combining multiple recall strategies, candidate products related to the user are screened out from different angles, thereby increasing the diversity and relevance of the recommended products; an interaction matrix and a weight matrix are constructed based on user data, the element values ​​of the elements in the interaction matrix represent the preference of each user for each candidate product, and the element values ​​of the elements in the weight matrix represent the weights of the corresponding elements in the interaction matrix; by applying a weighted matrix decomposition algorithm to the interaction matrix and the weight matrix, and then using user data and product data, user feature vectors and product feature vectors are calculated, which effectively captures the implicit features of users and products and improves the accuracy of recommendations; based on the obtained weight matrix, user feature vector and product feature vector, the matching degree between each user and each candidate product is calculated, and the target product of each user is determined based on the obtained matching degree, and the matching degree calculation makes the recommended products more in line with user needs; finally, personalized product recommendations are made based on the target product of each user, which is conducive to improving user satisfaction and click-through rate, and improving the accuracy of product recommendations.

[0119] In some optional implementations of this embodiment, the candidate determination module 302 may include: a first determination submodule, a second determination submodule, a third determination submodule, a fourth determination submodule, and a recall adjustment submodule, wherein:

[0120] The first determination submodule is used to determine the first recalled product in the candidate product library according to the product popularity data.

[0121] The second determination submodule is used to determine a second recalled product in the candidate product library according to the product activity data.

[0122] The third determination submodule is used to perform product-based collaborative filtering on each user to determine the third recalled product of each user in the candidate product library.

[0123] The fourth determination submodule is used to perform user-based collaborative filtering on each user to determine the fourth recalled product of each user in the candidate product library.

[0124] The recall adjustment submodule is used to perform recall adjustments on the recalled products of each user respectively to obtain the candidate products corresponding to each user.

[0125] In this embodiment, product popularity data and product activity data are combined, as well as product-based collaborative filtering and user-based collaborative filtering to provide users with richer and more personalized product recommendations, meet users' needs for popular products and promotional products, and meet users' potential interests, thereby ensuring the accuracy of subsequent product recommendations.

[0126] In some optional implementations of this embodiment, the vector calculation module 303 may include: a data extraction submodule, a factor acquisition submodule, an interaction construction submodule and a weight construction submodule, wherein:

[0127] The data extraction submodule is used to extract user behavior data from user data.

[0128] The factor acquisition submodule is used to obtain the interaction factor and interaction weight factor between each user and each candidate product based on user behavior data.

[0129] The interaction construction submodule is used to calculate the interaction score between each user and each candidate product according to the obtained interaction factor, and to construct an interaction matrix according to all the interaction scores.

[0130] The weight construction submodule is used to calculate the interaction weight between each user and each candidate product according to the obtained interaction weight factor, and to construct a weight matrix according to all the interaction weights.

[0131] In this embodiment, the interaction factors and interaction weight factors in the user behavior data are extracted to more comprehensively capture the interaction information between the user and the product; the interaction score between each user and each candidate product is calculated by weighted summing the interaction factors, reflecting the user's comprehensive interest in the product; the interaction intensity between the user and the product is reflected by weighted summing the interaction weight factors of different interaction behaviors; according to the interaction score and the interaction weight factor, the interaction matrix and the weight matrix are respectively constructed, which provide a data basis for the subsequent feature vector calculation and matching degree calculation.

[0132] In some optional implementations of this embodiment, the vector calculation module 303 may include: an initial generation submodule, an element multiplication submodule, a matrix decomposition submodule, a user calculation submodule, and a product calculation submodule, wherein:

[0133] The initial generation submodule is used to generate an initial user feature vector for each user based on the user data, and to generate an initial product feature vector for each candidate product based on the product data.

[0134] The element multiplication submodule is used to multiply the corresponding elements in the interaction matrix and the weight matrix one by one to obtain the weighted matrix.

[0135] The matrix decomposition submodule is used to perform matrix decomposition on the weighted matrix to obtain the user latent vector matrix and the product latent vector matrix.

[0136] The user calculation submodule is used to obtain the user latent vector of each user from the user latent vector matrix, and calculate the user feature vector of each user according to the user latent vector and the initial user feature vector.

[0137] The product calculation submodule is used to obtain the product latent vector of each candidate product from the product latent vector matrix, and calculate the product feature vector of each candidate product based on the product latent vector and the initial product feature vector.

[0138] In this embodiment, an initial user feature vector of the user is generated based on the user data, and an initial product feature vector of the candidate product is generated based on the product data, so as to capture the explicit features and basic attributes of the user and the product; the corresponding elements in the interaction matrix and the weight matrix are multiplied one by one to obtain a weighted matrix reflecting the user's weighted preference; the user latent vector matrix and the product latent vector matrix are extracted from the weighted matrix through a matrix decomposition algorithm, and the interaction information of the user and the product is compressed into a latent space to capture the potential features in the data; the user latent vector and the initial user feature vector are fused to calculate the user feature vector; the product latent vector and the initial product feature vector are fused to calculate the product feature vector, so as to ensure that the feature vector retains both the explicit features and the latent features, improve the accuracy of the feature vector, and ensure the accuracy of the subsequent matching degree calculation.

[0139] In some optional implementations of this embodiment, the product determination module 304 may include: a vector multiplication submodule, a weighted acquisition submodule, and a matching degree calculation submodule, wherein:

[0140] The vector multiplication submodule is used to multiply the user feature vector and the product feature vector to obtain the initial matching degree between each user and each candidate product.

[0141] The weighted acquisition submodule is used to obtain the weighted factor corresponding to the initial matching degree from the weight matrix.

[0142] The matching degree calculation submodule is used to multiply the weighted factor and the initial matching degree to obtain the matching degree between each user and each candidate product.

[0143] In this embodiment, the user feature vector and the product feature vector are multiplied to obtain the initial matching degree, thereby indicating the similarity between the user and the product in multiple dimensions, as well as the user's potential interest in the product; the corresponding weighting factor is then obtained from the weight matrix and multiplied with the initial matching degree, so that the final matching degree more accurately reflects the user's true preference, thereby improving the accuracy and personalization of the matching degree calculation.

[0144] In some optional implementations of this embodiment, the product determination module 304 may include: a descending sorting submodule, a first adjustment submodule, a second adjustment submodule, and a product determination submodule, wherein:

[0145] The descending sorting submodule is used to sort all candidate products in descending order according to the matching degree between each user and all candidate products to obtain a first product queue.

[0146] The first adjustment submodule is used to perform a first adjustment on the first product queue to obtain a second product queue according to the matching degree and the product label, wherein the first adjustment includes product screening and sorting adjustment.

[0147] The second adjustment submodule is used to receive the adjustment instruction sent by the control terminal, and perform a second adjustment on the second product queue according to the adjustment instruction to obtain a third product queue.

[0148] The product determination submodule is used to determine each product in the third product queue as a target product for each user.

[0149] In this embodiment, firstly, according to the matching degree between the user and all the candidate products, all the candidate products are sorted in descending order to obtain a first product queue; a first adjustment is made to the first product queue according to the matching degree and the product label to obtain an optimized second product queue; an adjustment instruction of the control terminal is received and responded to to obtain a third product queue so as to meet the promotion needs or platform strategies; through the optimization and adjustment in multiple dimensions, the relevance and personalization of the recommendation results are improved, so that the target products can better meet the needs of the users and improve the overall accuracy of the recommendations.

[0150] In some optional implementations of this embodiment, the product recommendation device 300 may include: a click acquisition module, a temporary determination module, and a temporary display module, wherein:

[0151] The click acquisition module is used to obtain each user's click information on the target product.

[0152] The temporary determination module is used to determine the temporary recommended products for each user based on the click information and the target products of each user.

[0153] The temporary display module is used to display temporary recommended products on the product page of each user's client.

[0154] In this embodiment, the user's click information on the target product is obtained to understand the user's latest interests and needs in real time; the user's click information and the target product are combined to dynamically generate temporary recommended products; the temporary recommended products are displayed on the user-side product page, allowing the user to browse and select new recommended products instantly, thereby improving the user's interactive experience and providing a more personalized and timely recommendation service.

[0155] 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.

[0156] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 having components 41-43, 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 (ASICs), field-programmable gate arrays (FPGAs), digital processors (DSPs), embedded devices, etc.

[0157] 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.

[0158] 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 equipped on the computer device 4, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. 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 computer-readable instructions of the product recommendation method, 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.

[0159] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip 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 product recommendation method.

[0160] 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.

[0161] The computer device provided in this embodiment can execute the above product recommendation method. Here, the product recommendation method can be the product recommendation method of each of the above embodiments.

[0162] In this embodiment, firstly, by combining multiple recall strategies, candidate products related to the user are screened out from different angles, thereby increasing the diversity and relevance of the recommended products; an interaction matrix and a weight matrix are constructed based on user data, the element values ​​of the elements in the interaction matrix represent the preference of each user for each candidate product, and the element values ​​of the elements in the weight matrix represent the weights of the corresponding elements in the interaction matrix; by applying a weighted matrix decomposition algorithm to the interaction matrix and the weight matrix, and then using user data and product data, user feature vectors and product feature vectors are calculated, which effectively captures the implicit features of users and products and improves the accuracy of recommendations; based on the obtained weight matrix, user feature vector and product feature vector, the matching degree between each user and each candidate product is calculated, and the target product of each user is determined based on the obtained matching degree, and the matching degree calculation makes the recommended products more in line with user needs; finally, personalized product recommendations are made based on the target product of each user, which is conducive to improving user satisfaction and click-through rate, and improving the accuracy of product recommendations.

[0163] The present application also provides another embodiment, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium 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 the product recommendation method as described above.

[0164] In this embodiment, firstly, by combining multiple recall strategies, candidate products related to the user are screened out from different angles, thereby increasing the diversity and relevance of the recommended products; an interaction matrix and a weight matrix are constructed based on user data, the element values ​​of the elements in the interaction matrix represent the preference of each user for each candidate product, and the element values ​​of the elements in the weight matrix represent the weights of the corresponding elements in the interaction matrix; by applying a weighted matrix decomposition algorithm to the interaction matrix and the weight matrix, and then using user data and product data, user feature vectors and product feature vectors are calculated, which effectively captures the implicit features of users and products and improves the accuracy of recommendations; based on the obtained weight matrix, user feature vector and product feature vector, the matching degree between each user and each candidate product is calculated, and the target product of each user is determined based on the obtained matching degree, and the matching degree calculation makes the recommended products more in line with user needs; finally, personalized product recommendations are made based on the target product of each user, which is conducive to improving user satisfaction and click-through rate, and improving the accuracy of product recommendations.

[0165] 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, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0166] 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. A product recommendation method, characterized in that: The steps include: Obtain a candidate product library, and obtain user data of all users and product data of all candidate products; According to a preset recall strategy, determining a candidate product corresponding to each user in the candidate product library; constructing an interaction matrix and a weight matrix according to the user data, and calculating a user feature vector of each user according to the interaction matrix, the weight matrix and the user data, and calculating a product feature vector of each candidate product according to the interaction matrix, the weight matrix and the product data; Calculate the matching degree between each user and each candidate product according to the obtained weight matrix, user feature vector and product feature vector, and determine the target product of each user based on the obtained matching degree; According to the target product of each user, product recommendations are made to each user.

2. The product recommendation method according to claim 1, characterized in that: The step of determining the candidate product corresponding to each user in the candidate product library according to the preset recall strategy includes: Determine a first recalled product in the candidate product library according to the product popularity data; determining a second recalled product in the candidate product library based on the product activity data; Performing product-based collaborative filtering on each user to determine a third recalled product of each user in the candidate product library; Performing user-based collaborative filtering on each user to determine the fourth recalled product of each user in the candidate product library; Recall adjustments are performed on the recalled products of each user respectively to obtain candidate products corresponding to each user.

3. The product recommendation method according to claim 1, characterized in that: The step of constructing an interaction matrix and a weight matrix according to the user data comprises: Extracting user behavior data from the user data; According to the user behavior data, obtaining an interaction factor and an interaction weight factor between each user and each candidate product; Calculating the interaction score between each user and each candidate product according to the obtained interaction factor, and constructing an interaction matrix according to all the interaction scores; The interaction weight between each user and each candidate product is calculated according to the obtained interaction weight factor, and a weight matrix is ​​constructed according to all the interaction weights.

4. The product recommendation method according to claim 3, characterized in that: The step of calculating the user feature vector of each user according to the interaction matrix, the weight matrix and the user data, and calculating the product feature vector of each candidate product according to the interaction matrix, the weight matrix and the product data comprises: Generating an initial user feature vector for each user according to the user data, and generating an initial product feature vector for each candidate product according to the product data; Multiplying the corresponding elements in the interaction matrix and the weight matrix one by one to obtain a weighted matrix; Performing matrix decomposition on the weighted matrix to obtain a user latent vector matrix and a product latent vector matrix; Acquire a user latent vector of each user from the user latent vector matrix, and calculate a user feature vector of each user according to the user latent vector and the initial user feature vector; The product latent vector of each candidate product is obtained from the product latent vector matrix, and the product feature vector of each candidate product is calculated according to the product latent vector and the initial product feature vector.

5. The product recommendation method according to claim 1, characterized in that: The step of calculating the matching degree between each user and each candidate product according to the obtained weight matrix, user feature vector and product feature vector comprises: Multiplying the user feature vector and the product feature vector to obtain an initial matching degree between each user and each candidate product; Acquire a weighting factor corresponding to the initial matching degree from the weight matrix; The weighting factor and the initial matching degree are multiplied to obtain the matching degree between each user and each candidate product.

6. The product recommendation method according to claim 1, characterized in that: The step of determining the target product of each user based on the obtained matching degree comprises: According to the matching degree between each user and all the candidate products, all the candidate products are sorted in descending order to obtain a first product queue; According to the matching degree and the product label, a first adjustment is performed on the first product queue to obtain a second product queue, wherein the first adjustment includes product screening and sorting adjustment; receiving an adjustment instruction sent by the control terminal, and performing a second adjustment on the second product queue according to the adjustment instruction to obtain a third product queue; Each product in the third product queue is determined as a target product for each user.

7. The product recommendation method according to claim 1, characterized in that: After the step of recommending products to each user according to the target product of each user, the method further includes: Obtaining click information of each user on the target product; Determine a temporary recommended product for each user according to the click information and the target product of each user; The temporarily recommended products are displayed on the product page of the user terminal held by each user.

8. A product recommendation device, characterized in that: include: An acquisition module is used to acquire a candidate product library and acquire user data of all users and product data of all candidate products; A candidate determination module, used to determine the candidate product corresponding to each user in the candidate product library according to a preset recall strategy; a vector calculation module, configured to construct an interaction matrix and a weight matrix according to the user data, and calculate a user feature vector of each user according to the interaction matrix, the weight matrix and the user data, and calculate a product feature vector of each candidate product according to the interaction matrix, the weight matrix and the product data; A product determination module, configured to calculate the matching degree between each user and each candidate product according to the obtained weight matrix, user feature vector and product feature vector, and determine the target product of each user based on the obtained matching degree; The product recommendation module is used to recommend products to each user according to the target products of each user.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the product recommendation method according to 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, and when the computer-readable instructions are executed by a processor, the steps of the product recommendation method according to any one of claims 1 to 7 are implemented.

Images