An e-commerce product recommendation method and device based on artificial intelligence

Abstract

The present invention discloses an e-commerce product recommendation method and device based on artificial intelligence, and relates to the field of artificial intelligence technology. The purchase record and the first recommendation model of the target user are called according to the terminal information; the target value matrix of the target user is calculated according to the purchase record, and the users whose similarity with the target value matrix is ​​greater than the first threshold are determined as the first user set; the price similarity and brand familiarity of the product are calculated, and the second user set is determined according to the price similarity and brand familiarity; the first recommendation model is corrected using the personal model parameters to obtain the recommendation probability of each product. The first user matching is performed using the value matrix, and the second user matching is performed according to the price similarity and brand similarity, and the second user matching is directly associated with the product itself, which reduces the influence of the user's individual purchasing habits on the recommendation, reduces the recommendation range, improves the accuracy and effectiveness of the recommendation, and enhances the user experience.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an e-commerce product recommendation method and device based on artificial intelligence. Background Art

[0002] As the storage capacity of computer technology increases, the amount of data generated by people is also growing exponentially. In this context, the traditional marketing model of enterprises is increasingly unable to adapt to the complex and changing social environment, especially the e-commerce industry relying on the Internet. In addition, with the huge number of customers, it is becoming increasingly difficult to carry out marketing activities for customers. Moreover, with the continuous improvement of the online shopping system, consumers' consumption psychology and consumption forms are changing, and traditional marketing methods are far from meeting the needs of today's consumers. At this stage, many e-commerce companies do not have clear target customers in marketing, and carry out large-scale, full-coverage publicity and promotion for all users, which will cause users to receive a large number of messages without any difference. In the long run, it may cause users to resist and lead to the loss of target users.

[0003] When customers interact with products on shopping websites or apps, they leave interactive information such as browsing history, purchase history, and comments about the products. Using this interactive information helps to reach target customers. Many studies have studied machine learning models to build recommendation systems. The common point of recommendation systems is to filter out the products that are most relevant to customers so that they can be retained for a longer period of time and improve the customer's product experience.

[0004] The most widely used and mature recommendation algorithm is the collaborative filtering (CF) recommendation algorithm. The basic idea of ​​CF is to recommend items to users based on their previous preferences and the choices of other users with similar interests. However, the CF recommendation algorithm in the prior art relies on other users with similar interests rather than products, and the individual purchasing habits of each user will make the recommendation range too large, resulting in low accuracy and effectiveness of the recommendation, which greatly affects the user experience. Summary of the invention

[0005] The purpose of the present invention is to solve the problems of the above-mentioned background technology and to propose an e-commerce product recommendation method and device based on artificial intelligence.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] An embodiment of the present invention provides an e-commerce product recommendation method based on artificial intelligence, which is applied to a cloud server and includes:

[0008] Acquire terminal information sent by a user terminal corresponding to a target user, and call the purchase record of the target user and a first recommendation model according to the terminal information; the terminal information includes a user tag and a personal model parameter of the target user; the first recommendation model is an average model of the target user's association network;

[0009] Calculate the target value matrix of the target user according to the purchase record, match the value matrices of other users, and determine users whose similarity with the target value matrix is ​​greater than a first threshold as a first user set;

[0010] For each user in the first user set, calculate the price similarity and brand similarity of the relevant products between the user and the target user, and determine the users whose weighted sum of price similarity and brand similarity is greater than a second threshold as the second user set;

[0011] Using the personal model parameters to modify the first recommendation model, and using the relevant products of each user in the second user set as model input to obtain the recommendation probability of each product;

[0012] The product information is returned to the user terminal as a first e-commerce product recommendation display list according to the recommendation probability.

[0013] Optionally, before obtaining terminal information sent by a user terminal corresponding to the target user and calling the purchase record and recommendation model of the target user according to the terminal information, the method further includes:

[0014] When it is detected that the target user logs in for the first time, a user tag of the target user is created, and a social network of the target user is obtained as an associated network; each user in the associated network is used as a node, and users are connected based on trust probability as connection edges between nodes, and the trust probability between the target user and any user in the associated network is higher than a third threshold;

[0015] Obtaining the purchase records of each user in the association network and inputting them into a preset recommendation model to obtain the recommendation probability of each product, and returning product information to the user terminal as a second e-commerce product recommendation display list according to the recommendation probability;

[0016] Modifying the preset recommendation model according to the browsing and purchasing results returned by the user terminal to obtain a second recommendation model;

[0017] The model parameters of the second recommendation model are returned to the user terminal as personal model parameters for storage, and the model parameters corresponding to each user in the association relationship network are averaged to obtain the first recommendation model.

[0018] Optionally, the number of users in the associated relationship network is a fourth threshold; and the second user set includes the associated relationship network.

[0019] Optionally, calculating the target value matrix of the target user according to the purchase record includes:

[0020] Preprocessing the purchase record to remove abnormal data and perform data standardization to obtain data to be processed;

[0021] Calculate the purchase frequency, average purchase quantity, average expenditure and purchase conversion rate of the target user according to the data to be processed;

[0022] The covariance matrix composed of the purchase frequency, the average purchase quantity, the average expenditure and the purchase conversion rate is used as the target value matrix of the target user.

[0023] Optionally, calculating the price similarity and brand similarity of related products between the user and the target user includes:

[0024] For any related product between the user and the target user, calculate the relative price of the user for the related product:

[0025]

[0026] Among them, X i is the relative price of the i-th related product, Z is the sum of the prices of all products purchased by the user, and s j Indicates the price of any product purchased by the user, S i is the actual price of the i-th related product;

[0027] Determine the total price of all products of the category corresponding to the relevant product purchased by the user, and determine the ratio of the total price of the relevant product to the total price as the relative brand value of the brand of the relevant product;

[0028] Calculate the price similarity of the relevant product based on the relative price of the relevant product between the target user and the user;

[0029] The brand similarity of the relevant product is calculated based on the relative brand value of the target user and the user regarding the relevant product.

[0030] A second aspect of an embodiment of the present invention provides an e-commerce product recommendation device based on artificial intelligence, the device being applied to a cloud server, comprising:

[0031] A first data acquisition module is used to acquire terminal information sent by a user terminal corresponding to a target user, and call the purchase record of the target user and a first recommendation model according to the terminal information; the terminal information includes a user tag and a personal model parameter of the target user; the first recommendation model is an average model of the target user's associated relationship network;

[0032] A first user set module, configured to calculate a target value matrix of the target user according to the purchase record, match the value matrices of other users, and determine users whose similarity with the target value matrix is ​​greater than a first threshold as a first user set;

[0033] A second user set module, configured to calculate, for each user in the first user set, price similarity and brand similarity of related products between the user and the target user, and determine users whose weighted sum of price similarity and brand similarity is greater than a second threshold as a second user set;

[0034] A recommendation calculation module, used to modify the first recommendation model using personal model parameters, and use the relevant products of each user in the second user set as model input to obtain the recommendation probability of each product;

[0035] The first recommendation returning module is used to return product information to the user terminal as a first e-commerce product recommendation display list according to the recommendation probability.

[0036] Optionally, the device further comprises:

[0037] A second data acquisition module is used to create a user tag of the target user when detecting that the target user logs in for the first time, and acquire the social relationship network of the target user as an associated relationship network; each user in the associated relationship network is used as a node, and the users are connected based on the trust probability as the connection edge between the nodes, and the trust probability between the target user and any user in the associated relationship network is higher than a third threshold;

[0038] A second recommendation returning module is used to obtain the purchase records of each user in the association relationship network and input them into a preset recommendation model to obtain the recommendation probability of each product, and return product information to the user terminal as a second e-commerce product recommendation display list according to the recommendation probability;

[0039] A model correction module, used to correct the preset recommendation model according to the browsing and purchasing results returned by the user terminal to obtain a second recommendation model;

[0040] The model storage module returns the model parameters of the second recommendation model as personal model parameters to the user terminal for storage, and averages the model parameters corresponding to each user in the association network to obtain the first recommendation model.

[0041] Optionally, the number of users in the associated relationship network is a fourth threshold; and the second user set includes the associated relationship network.

[0042] Optionally, the first user set module includes a target value matrix module; the target value matrix module includes:

[0043] A preprocessing module, used for preprocessing the purchase record line to remove abnormal data and perform data standardization to obtain data to be processed;

[0044] A feature calculation module calculates the purchase frequency, average purchase quantity, average expenditure and purchase conversion rate of the target user according to the data to be processed;

[0045] A matrix construction module is used to use the covariance matrix composed of the purchase frequency, the average purchase quantity, the average expenditure and the purchase conversion rate as the target value matrix of the target user.

[0046] Optionally, the second user set module includes a familiarity module, and the familiarity module includes:

[0047] The relative price calculation module is used to calculate the relative price of the user with respect to any related product between the user and the target user:

[0048]

[0049] Among them, X i is the relative price of the i-th related product, Z is the sum of the prices of all products purchased by the user, and s j Indicates the price of any product purchased by the user, S i is the actual price of the i-th related product;

[0050] A relative brand value calculation module is used to determine the total price of all products of the category corresponding to the relevant product purchased by the user, and determine the ratio of the total price of the relevant product to the total price as the relative brand value of the brand of the relevant product;

[0051] A first similarity module, calculating the price similarity of the related product according to the relative price of the related product between the target user and the user;

[0052] The second similarity module calculates the brand similarity of the related product according to the relative brand value of the target user and the user regarding the related product.

[0053] Beneficial effects of the present invention:

[0054] The embodiment of the present invention provides an e-commerce product recommendation method based on artificial intelligence, which is applied to a cloud server and includes: obtaining terminal information sent by a user terminal corresponding to a target user, calling the target user's purchase record and a first recommendation model according to the terminal information; the terminal information includes a user tag and a personal model parameter of the target user; the first recommendation model is an average model of the target user's association network; calculating a target value matrix of the target user according to the purchase record, matching the value matrices of other users, and determining users whose similarity with the target value matrix is ​​greater than a first threshold as a first user set; for each user in the first user set, calculating the price similarity and brand similarity of related products between the user and the target user, and determining users whose weighted sum value of price similarity and brand similarity is greater than a second threshold as a second user set; using the personal model parameters to correct the first recommendation model, and using the related products of each user in the second user set as model input to obtain the recommendation probability of each product; and returning product information to the user terminal according to the recommendation probability as a first e-commerce product recommendation display list. The value matrix is ​​used for the first user matching, and the second user matching is performed based on price similarity and brand similarity. The second user matching is directly related to the product itself, which reduces the impact of the user's individual purchasing habits on the recommendation, reduces the recommendation scope, improves the accuracy and effectiveness of the recommendation, and enhances the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The present invention will be further described below in conjunction with the accompanying drawings.

[0056] Figure 1 A flowchart of an e-commerce product recommendation method based on artificial intelligence provided by an embodiment of the present invention;

[0057] Figure 2 A structural diagram of an e-commerce product recommendation device based on artificial intelligence provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] The embodiment of the present invention provides an e-commerce product recommendation method based on artificial intelligence, see Figure 1 , Figure 1 A flowchart of an e-commerce product recommendation method based on artificial intelligence provided by an embodiment of the present invention, the method is applied to a cloud server, comprising:

[0060] S1, obtaining terminal information sent by a user terminal corresponding to a target user, and calling a purchase record and a first recommendation model of the target user according to the terminal information.

[0061] S2, calculating the target value matrix of the target user according to the purchase records, matching the value matrices of other users, and determining the users whose similarity with the target value matrix is ​​greater than a first threshold as the first user set.

[0062] S3, for each user in the first user set, calculate the price similarity and brand similarity of the relevant products between the user and the target user, and determine the users whose weighted sum of price similarity and brand similarity is greater than a second threshold as the second user set.

[0063] S4, using the personal model parameters to modify the first recommendation model, and using the relevant products of each user in the second user set as model input to obtain the recommendation probability of each product.

[0064] S5: Return product information to the user terminal as a first e-commerce product recommendation display list according to the recommendation probability.

[0065] The terminal information includes the user tag and the personal model parameters of the target user; the first recommended model is the average model of the target user's associated relationship network.

[0066] An artificial intelligence-based e-commerce product recommendation method provided by an embodiment of the present invention uses a value matrix to perform a first user match, and performs a second user match based on price similarity and brand similarity. The second user match is directly associated with the product itself, which reduces the impact of the user's individual purchasing habits on the recommendation, reduces the recommendation scope, improves the accuracy and effectiveness of the recommendation, and enhances the user experience.

[0067] In one embodiment, before step S1, the method further comprises:

[0068] Step 1: when it is detected that the target user logs in for the first time, a user tag of the target user is created, and a social network of the target user is obtained as an associated network; each user in the associated network is used as a node, and the users are connected based on the trust probability as the connection edge between the nodes, and the trust probability between the target user and any user in the associated network is higher than a third threshold;

[0069] Step 2: Obtain the purchase records of each user in the association network and input them into a preset recommendation model to obtain the recommendation probability of each product, and return product information to the user terminal as a second e-commerce product recommendation display list according to the recommendation probability;

[0070] Step 3: modifying the preset recommendation model according to the browsing and purchasing results returned by the user terminal to obtain a second recommendation model;

[0071] Step 4: Return the model parameters of the second recommendation model as personal model parameters to the user terminal for storage, and average the model parameters corresponding to each user in the association network to obtain the first recommendation model.

[0072] In one implementation, when the target user logs in for the first time, there is no browsing history or purchase history, resulting in no direction for recommendation. Building an associated relationship network through a social relationship network can solve the cold start problem of the recommendation system. The preset recommendation model can be any e-commerce product recommendation model in the prior art, and is not limited here. The trust probability between users can be obtained based on the communication frequency between users.

[0073] In one implementation, the cloud server only saves the first recommended model, and the user terminal saves the personal model parameters. The cloud server obtains the personal model parameters only when the user uses it, which can ensure the privacy of personal data.

[0074] In one embodiment, the number of users in the associated relationship network is a fourth threshold; and the second user set includes the associated relationship network.

[0075] In one implementation, users of the association network can be replaced according to the weighted sum of price similarity and brand similarity, so that users with large weighted sum values ​​are always kept in the association network, thereby ensuring a dynamic improvement in accuracy.

[0076] In one embodiment, calculating the target value matrix of the target user according to the purchase record includes:

[0077] Step 1: pre-process the purchase record to remove abnormal data and perform data standardization to obtain the data to be processed;

[0078] Step 2: Calculate the target user's purchase frequency, average purchase quantity, average spending and purchase conversion rate based on the data to be processed;

[0079] Step three: Use the covariance matrix composed of purchase frequency, average purchase quantity, average spending and purchase conversion rate as the target value matrix of target users.

[0080] In one embodiment, calculating the price similarity and brand similarity of related products between the user and the target user includes:

[0081] Step 1: For any related product between the user and the target user, calculate the relative price of the user for the related product:

[0082]

[0083] Among them, X i is the relative price of the i-th related product, Z is the sum of the prices of all products purchased by the user, and s jIndicates the price of any product purchased by the user, S i is the actual price of the i-th related product;

[0084] Step 2: Determine the total price of all products in the category corresponding to the relevant product purchased by the user, and determine the ratio of the total price of the relevant product to the total price, which serves as the relative brand value of the brand of the relevant product;

[0085] Step 3, calculating the price similarity of the relevant product based on the relative price of the relevant product between the target user and the user;

[0086] Step 4: Calculate the brand similarity of the relevant product based on the relative brand value of the target user and the user regarding the relevant product.

[0087] The embodiment of the present invention provides an e-commerce product recommendation device based on artificial intelligence, see Figure 2 , Figure 2 A structural diagram of an e-commerce product recommendation device based on artificial intelligence provided by an embodiment of the present invention, the device is applied to a cloud server, including:

[0088] The first data acquisition module is used to acquire terminal information sent by the user terminal corresponding to the target user, and call the purchase record of the target user and the first recommendation model according to the terminal information; the terminal information includes the user tag and the personal model parameters of the target user; the first recommendation model is the average model of the target user's associated relationship network;

[0089] A first user set module, configured to calculate a target value matrix of a target user according to the purchase records, match the value matrices of other users, and determine users whose similarity to the target value matrix is ​​greater than a first threshold as a first user set;

[0090] A second user set module is used to calculate, for each user in the first user set, the price similarity and brand similarity of the relevant products between the user and the target user, and determine the users whose weighted sum of the price similarity and the brand similarity is greater than a second threshold as the second user set;

[0091] A recommendation calculation module, used to modify the first recommendation model using the personal model parameters, and use the relevant products of each user in the second user set as model input to obtain the recommendation probability of each product;

[0092] The first recommendation returning module is used to return product information to the user terminal as a first e-commerce product recommendation display list according to the recommendation probability.

[0093] An artificial intelligence-based e-commerce product recommendation device provided based on an embodiment of the present invention uses a value matrix to perform a first user match, and performs a second user match based on price similarity and brand similarity, and the second user match is directly associated with the product itself, thereby reducing the impact of the user's individual purchasing habits on the recommendation, reducing the recommendation scope, improving the accuracy and effectiveness of the recommendation, and enhancing the user experience.

[0094] In one embodiment, the apparatus further comprises:

[0095] The second data acquisition module is used to create a user tag of the target user when detecting that the target user logs in for the first time, and obtain the social relationship network of the target user as an associated relationship network; each user in the associated relationship network is used as a node, and the users are connected based on the trust probability as the connection edge between the nodes, and the trust probability between the target user and any user in the associated relationship network is higher than a third threshold;

[0096] The second recommendation return module is used to obtain the purchase records of each user in the association network and input them into a preset recommendation model to obtain the recommendation probability of each product, and return product information to the user terminal as a second e-commerce product recommendation display list according to the recommendation probability;

[0097] A model correction module, used to correct the preset recommendation model according to the browsing and purchasing results returned by the user terminal to obtain a second recommendation model;

[0098] The model storage module returns the model parameters of the second recommendation model as personal model parameters to the user terminal for storage, and averages the model parameters corresponding to each user in the association network to obtain the first recommendation model.

[0099] In one embodiment, the number of users in the associated relationship network is a fourth threshold; and the second user set includes the associated relationship network.

[0100] In one embodiment, the first user set module includes a target value matrix module; the target value matrix module includes:

[0101] A preprocessing module is used to preprocess the purchase record lines to remove abnormal data and perform data standardization to obtain data to be processed;

[0102] The feature calculation module calculates the target user's purchase frequency, average purchase quantity, average spending and purchase conversion rate based on the data to be processed;

[0103] The matrix building module is used to use the covariance matrix composed of purchase frequency, average purchase quantity, average expenditure and purchase conversion rate as the target value matrix of the target user.

[0104] In one embodiment, the second user set module includes a familiarity module, and the familiarity module includes:

[0105] The relative price calculation module is used to calculate the relative price of the user for any related product between the user and the target user:

[0106]

[0107] Among them, X i is the relative price of the i-th related product, Z is the sum of the prices of all products purchased by the user, and s j Indicates the price of any product purchased by the user, S i is the actual price of the i-th related product;

[0108] A relative brand value calculation module is used to determine the total price of all products of the category corresponding to the relevant product purchased by the user, and determine the ratio of the total price of the relevant product to the total price as the relative brand value of the brand of the relevant product;

[0109] A first similarity module, calculating the price similarity of the related product according to the relative price of the related product between the target user and the user;

[0110] The second similarity module calculates the brand similarity of the related product according to the relative brand value of the target user and the user regarding the related product.

[0111] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state hard disk Solid State Disk (SSD)), etc.

[0112] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0113] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system, electronic device and storage medium, since they are basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0114] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. An e-commerce product recommendation method based on artificial intelligence, characterized in that: The method is applied to a cloud server, comprising: Acquire terminal information sent by a user terminal corresponding to a target user, and call the purchase record of the target user and a first recommendation model according to the terminal information; the terminal information includes a user tag and a personal model parameter of the target user; the first recommendation model is an average model of the target user's association network; Calculate the target value matrix of the target user according to the purchase record, match the value matrices of other users, and determine users whose similarity with the target value matrix is ​​greater than a first threshold as a first user set; For each user in the first user set, calculate the price similarity and brand similarity of the relevant products between the user and the target user, and determine the users whose weighted sum of price similarity and brand similarity is greater than a second threshold as the second user set; Using the personal model parameters to modify the first recommendation model, and using the relevant products of each user in the second user set as model input to obtain the recommendation probability of each product; Returning product information to the user terminal as a first e-commerce product recommendation display list according to the recommendation probability; Before obtaining terminal information sent by a user terminal corresponding to the target user and calling the purchase record and the first recommendation model of the target user according to the terminal information, the method further includes: When it is detected that the target user logs in for the first time, a user tag of the target user is created, and a social network of the target user is obtained as an associated network; each user in the associated network is used as a node, and users are connected based on trust probability as connection edges between nodes, and the trust probability between the target user and any user in the associated network is higher than a third threshold; Obtaining the purchase records of each user in the association network and inputting them into a preset recommendation model to obtain the recommendation probability of each product, and returning product information to the user terminal as a second e-commerce product recommendation display list according to the recommendation probability; Modifying the preset recommendation model according to the browsing and purchasing results returned by the user terminal to obtain a second recommendation model; Returning the model parameters of the second recommendation model as personal model parameters to the user terminal for storage, and averaging the model parameters corresponding to each user in the association network to obtain the first recommendation model; Calculating the target value matrix of the target user according to the purchase record includes: Preprocessing the purchase record to remove abnormal data and perform data standardization to obtain data to be processed; Calculate the purchase frequency, average purchase quantity, average expenditure and purchase conversion rate of the target user according to the data to be processed; The covariance matrix composed of the purchase frequency, the average purchase quantity, the average expenditure and the purchase conversion rate is used as the target value matrix of the target user.

2. The method for recommending e-commerce products based on artificial intelligence according to claim 1, characterized in that: The number of users in the associated relationship network is a fourth threshold; and the second user set includes the associated relationship network.

3. The method for recommending e-commerce products based on artificial intelligence according to claim 1, characterized in that: Calculating the price similarity and brand similarity of the relevant products between the user and the target user includes: For any related product between the user and the target user, calculate the relative price of the user for the related product: Among them, X i is the relative price of the i-th related product, Z is the sum of the prices of all products purchased by the user, and s j Indicates the price of any product purchased by the user, S i is the actual price of the i-th related product; Determine the total price of all products of the category corresponding to the relevant product purchased by the user, and determine the ratio of the total price of the relevant product to the total price as the relative brand value of the brand of the relevant product; Calculate the price similarity of the relevant product based on the relative price of the relevant product between the target user and the user; The brand similarity of the relevant product is calculated based on the relative brand value of the target user and the user regarding the relevant product.

4. An e-commerce product recommendation device based on artificial intelligence, characterized in that: The device is applied to a cloud server and includes: A first data acquisition module is used to acquire terminal information sent by a user terminal corresponding to a target user, and call the purchase record of the target user and a first recommendation model according to the terminal information; the terminal information includes a user tag and a personal model parameter of the target user; the first recommendation model is an average model of the target user's associated relationship network; A first user set module, configured to calculate a target value matrix of the target user according to the purchase record, match the value matrices of other users, and determine users whose similarity with the target value matrix is ​​greater than a first threshold as a first user set; A second user set module, configured to calculate, for each user in the first user set, price similarity and brand similarity of related products between the user and the target user, and determine users whose weighted sum of price similarity and brand similarity is greater than a second threshold as a second user set; A recommendation calculation module, used to modify the first recommendation model using personal model parameters, and use the relevant products of each user in the second user set as model input to obtain the recommendation probability of each product; A first recommendation returning module, configured to return product information to the user terminal as a first e-commerce product recommendation display list according to the recommendation probability; The device also includes: A second data acquisition module is used to create a user tag of the target user and acquire a social network of the target user as an associated network when detecting that the target user logs in for the first time; Each user in the association network is a node, and users are connected based on trust probability as connection edges between nodes, and the trust probability between the target user and any user in the association network is higher than a third threshold; A second recommendation returning module is used to obtain the purchase records of each user in the association relationship network and input them into a preset recommendation model to obtain the recommendation probability of each product, and return product information to the user terminal as a second e-commerce product recommendation display list according to the recommendation probability; A model correction module, used to correct the preset recommendation model according to the browsing and purchasing results returned by the user terminal to obtain a second recommendation model; A model storage module, returning the model parameters of the second recommendation model as personal model parameters to the user terminal for storage, and averaging the model parameters corresponding to each user in the association network to obtain the first recommendation model; The first user set module includes a target value matrix module; The target value matrix module includes: A preprocessing module, used for preprocessing the purchase record line to remove abnormal data and perform data standardization to obtain data to be processed; A feature calculation module calculates the purchase frequency, average purchase quantity, average expenditure and purchase conversion rate of the target user according to the data to be processed; A matrix construction module is used to use the covariance matrix composed of the purchase frequency, the average purchase quantity, the average expenditure and the purchase conversion rate as the target value matrix of the target user.

5. The artificial intelligence-based e-commerce product recommendation device according to claim 4, characterized in that: The number of users in the associated relationship network is a fourth threshold; and the second user set includes the associated relationship network.

6. The artificial intelligence-based e-commerce product recommendation device according to claim 4, characterized in that: The second user set module includes a familiarity module, and the familiarity module includes: The relative price calculation module is used to calculate the relative price of the user with respect to any related product between the user and the target user: Among them, X i is the relative price of the i-th related product, Z is the sum of the prices of all products purchased by the user, and s j Indicates the price of any product purchased by the user, S i is the actual price of the i-th related product; A relative brand value calculation module is used to determine the total price of all products of the category corresponding to the relevant product purchased by the user, and determine the ratio of the total price of the relevant product to the total price as the relative brand value of the brand of the relevant product; A first similarity module, calculating the price similarity of the related product according to the relative price of the related product between the target user and the user; The second similarity module calculates the brand similarity of the related product according to the relative brand value of the target user and the user regarding the related product.

Images