An intelligent recommendation system for website interface information based on big data analysis

Abstract

The present invention provides a website interface information intelligent recommendation system based on big data analysis, which relates to the technical field of website information recommendation, including: clustering user portraits to obtain multiple user groups; determining the target user group based on the target user; obtaining the historical information browsing sequence and the historical information recommendation sequence of the target user group; obtaining the current information browsing sequence of the target user, and generating multiple groups of current information recommendation sequences based on the current information browsing sequence; the information browsing sequence includes a website, browsing time and browsing action; based on the historical information browsing sequence and the historical information recommendation sequence, scoring the current information recommendation sequence, and recommending information to the target user according to the score; the score indicates the degree to which the current information recommendation sequence meets the information browsing preference of the target user.

Description

Technical Field

[0001] The present invention relates to the technical field of website information recommendation, and in particular to a website interface information intelligent recommendation system based on big data analysis. Background Art

[0002] With the development of network technology, the information on the Internet is becoming increasingly complex. Users need to search for the information they need in a large amount of useless information, which greatly consumes the energy and time of users. In order to reduce the search time of users, websites have launched website information recommendation systems. However, the existing website information recommendation systems only recommend information based on the user's discrete historical behavior. However, during the process of information search and absorption, users' preferences often shift to a certain extent, making the website's recommended information no longer suitable for the user's current needs.

[0003] In view of this, the present invention proposes a website interface information intelligent recommendation system based on big data analysis, which obtains characteristic information related to changes in user preference by processing the information browsing sequence and information recommendation sequence of the user, and provides information recommendation services to the user based on the characteristic information, thereby improving the accuracy of website information recommendation. Summary of the invention

[0004] The present invention aims to provide a website interface information intelligent recommendation system based on big data analysis, comprising: clustering user portraits to obtain multiple user groups; determining a target user group based on a target user; obtaining a historical information browsing sequence and a historical information recommendation sequence of the target user group; obtaining a current information browsing sequence of the target user, and generating multiple groups of current information recommendation sequences based on the current information browsing sequence; the information browsing sequence includes a website, browsing time and browsing action; scoring the current information recommendation sequence based on the historical information browsing sequence and the historical information recommendation sequence, and recommending information to the target user according to the score; the score indicates the degree to which the current information recommendation sequence meets the information browsing preference of the target user; wherein scoring the current information recommendation sequence comprises: obtaining a second user similarity between the target user and other users in the target user group; determining the current group score of other users for the current information recommendation sequence and the group average score for the corresponding historical information recommendation sequence; determining the target average score of the target user for the target historical information recommendation sequence; and determining the score of the current information recommendation sequence based on the second user similarity, the current group score, the group average score and the target average score.

[0005] Furthermore, clustering the user portraits to obtain multiple user groups includes: constructing a first user feature vector based on the user portraits; determining a first user similarity by calculating a distance between the first user feature vectors; determining a user weight based on the first user feature vector and the first user similarity; concatenating the first user feature vector and the user weight to obtain a second user feature vector; and classifying users based on the total information category and the second user feature vector to obtain the multiple user groups.

[0006] Furthermore, the expression of the first user similarity is: ;in, represents the value of the first user similarity between user n and user i, and Respectively represent different first user feature vectors, n and i represent user variables; m represents the first feature variable; M represents the total number of first features; Indicates the importance of the mth first feature; Represents the first feature and the first feature The distance between represents the mth first feature of the nth user; represents the mth first feature of the i-th user.

[0007] Furthermore, the expression for determining the user weight is: ;in, Indicates the user weight of the nth user; Indicates taking the absolute value; Represents the first user feature vector of the nth user The average distance to other users’ feature vectors; Represents the first user feature vector of the nth user The maximum distance to other user feature vectors; Represents the first user feature vector of the nth user The minimum value of the distance to other user feature vectors; m represents the first feature variable; M represents the total number of first features; Indicates the importance of the mth first feature; represents the mth first feature of the nth user; represents the mean of the mth first feature of all users; represents the maximum value of the mth first feature; represents the minimum value of the mth first feature; represents the L2 paradigm; represents the first user feature vector of the i-th user; represents the first user feature vector of the kth user; U represents the first user feature vector set ; Represents the first user feature vector of the i-th user The average distance to other users’ feature vectors; Represents the first user feature vector of the i-th user The maximum distance to other user feature vectors; Represents the first user feature vector of the i-th user The minimum distance to other user feature vectors; represents the value of the first user similarity between user n and user i; represents the value of the first user similarity between user i and user k; represents the user weight of the i-th user.

[0008] Furthermore, the classifying of users to obtain the multiple user groups includes: determining the total number of information categories; determining the second user similarity based on the total number of information categories and the user weights; the expression of the second user similarity is: ;in, Represents the second user similarity value between user n and user i; A second user feature vector representing user n; represents the second user feature vector of user i; j represents the commodity variable; C represents the total number of commodity categories; Feature weight coefficient representing information; The absolute value is taken; the users whose similarity degree to the second user is greater than the preset similarity threshold are divided into a user group. The characteristic weight coefficient of the information can be set according to the actual needs of the website.

[0009] Furthermore, the method of generating multiple groups of current information recommendation sequences based on the current information browsing sequence includes: obtaining the title and image of the website page based on the URL in the current information browsing sequence; inputting the title into a word vector model to obtain a title vector sequence; inputting the image into a graph convolutional neural network to obtain an image vector sequence; concatenating the title vector sequence and the image vector sequence to obtain a current browsing vector sequence; inputting the current browsing vector sequence into a long short-term memory network to obtain a current browsing sequence hidden state; concatenating the current browsing sequence hidden state and a second user feature vector to obtain a current user feature vector; matching the current user feature vector with the information feature vector to obtain the multiple groups of current information recommendation sequences.

[0010] Furthermore, the information feature vector is obtained by performing representation learning on the information knowledge graph, including: using the transE algorithm to perform representation learning on the information knowledge graph to obtain an entity representation matrix E and a relationship representation matrix R; extracting information from the title to obtain interest information of the target user; based on the entity representation matrix and the relationship representation matrix of the interest information, establishing an information feature vector related to the user's interest; the expression of the information feature vector is: ;in, represents the vth information feature vector; e represents the entity corresponding to the vth information, which refers to the information in the website; E represents the entity representation matrix; Indicate interest information; Shows the distance between information e and interest information in the information knowledge graph; Indicates the distance between the interest information and other information in the information knowledge graph; Indicates other information than information e; Represents the embedding of the vth information.

[0011] Furthermore, obtaining the multiple groups of current information recommendation sequences includes: using a multi-head attention mechanism to process the current user feature vector and the information feature vector to obtain multiple initial attention matrices; using a cross-attention mechanism to process the multiple initial attention matrices to obtain a cross-attention matrix; inputting the cross-attention matrix into a fully connected layer to obtain a matching probability between the target user and the information; using the information with a matching probability greater than a preset matching threshold as recommendation information for the multiple groups of current information recommendation sequences; constructing a predicted browsing sequence based on the recommendation information, using the predicted browsing sequence as a new current information browsing sequence for calculating new recommendation information, and sorting them in the chronological order of obtaining the recommended information to obtain multiple groups of current information recommendation sequences.

[0012] Furthermore, the score of the current information recommendation sequence is expressed as: ;in, Indicates the score of the current information recommendation sequence; represents the target average rating; User variables representing the target user group; Represents the total number of users in the target user group; Indicates users in the target user group With target users The similarity of the second user; Indicates the difference between the current group score and the group average score.

[0013] Furthermore, the website is a shopping website; the information is commodities; the browsing actions include clicking, adding to cart and placing an order; and the commodity recommendation sequence is scored by analyzing the user's browsing actions on the recommended commodities.

[0014] The technical solution of the embodiment of the present invention has at least the following advantages and beneficial effects:

[0015] The present invention optimizes the current information recommendation sequence of a website to a target user through a target user group, and continuously updates the subsequent information recommendation sequence based on user feedback, thereby improving the accuracy of information pushed by the website.

[0016] The present invention constructs a first user feature vector, constructs a second user feature vector based on the first user feature vector, and classifies users based on the information category of the website, so that the users of the website can be accurately classified based on the characteristics of the website information, so that the classification of users is more consistent with the content of the website, which is convenient for subsequent recommendation operations.

[0017] The present invention obtains a hidden state related to the change of user selection preference by inputting the current browsing vector sequence into the long short-term memory network, and matches the current information recommendation sequence for the user based on the hidden state and the user feature vector, thereby further improving the accuracy of the current information recommendation sequence.

[0018] The present invention predicts the target user's score for the current information recommendation sequence through the similarity between the target user group and the target user and their score for the current information recommendation sequence, and can recommend the current information recommendation sequence that better meets the user's preferences to the user, thereby avoiding a large difference between the recommended information and the user's preferences. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 An exemplary flow chart of a website interface information intelligent recommendation system based on big data analysis provided by the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0021] Figure 1 This is an exemplary flow chart of a website interface information intelligent recommendation system based on big data analysis provided by the present invention. Figure 1 As shown, the process of the website interface information intelligent recommendation system includes the following steps:

[0022] Step 1: Cluster the user portraits to obtain multiple user groups. The user portrait includes the user's gender, age, address, consumption level, hobbies, etc. A user group refers to a set of multiple users obtained by classifying users. For example, a website can be a shopping website, and a user can refer to a consumer who shops on a shopping website. The user portrait can include the user's gender, age, delivery address, consumption level, hobbies, etc. By clustering users, users can be divided into multiple user groups with similar consumption levels and similar ages.

[0023] Cluster the user portraits to obtain multiple user groups, including:

[0024] Based on the user portrait , where the user variable is and , construct the first user feature vector , where the first characteristic variable is and .

[0025] The first user similarity is determined by calculating the distance between the first user feature vectors. The expression of the first user similarity is:

[0026] ;

[0027] in, represents the value of the first user similarity between user n and user i, and Respectively represent different first user feature vectors, n and i represent user variables; m represents the first feature variable; M represents the total number of first features; Indicates the importance of the mth first feature; Represents the first feature and the first feature The distance between represents the mth first feature of the nth user; represents the mth first feature of the i-th user.

[0028] Determine a user weight based on the first user feature vector and the first user similarity The expression for determining the user weight is:

[0029] ;

[0030] in, Indicates the user weight of the nth user; Indicates taking the absolute value; Represents the first user feature vector of the nth user The average distance to other users’ feature vectors; Represents the first user feature vector of the nth user The maximum distance to other user feature vectors; Represents the first user feature vector of the nth user The minimum value of the distance to other user feature vectors; m represents the first feature variable; M represents the total number of first features; Indicates the importance of the mth first feature; represents the mth first feature of the nth user; represents the mean of the mth first feature of all users; represents the maximum value of the mth first feature; represents the minimum value of the mth first feature; represents the L2 paradigm; represents the first user feature vector of the i-th user; represents the first user feature vector of the kth user; U represents the first user feature vector set ; Represents the first user feature vector of the i-th user The average distance to other users’ feature vectors; Represents the first user feature vector of the i-th user The maximum distance to other user feature vectors; Represents the first user feature vector of the i-th user The minimum distance to other user feature vectors; represents the value of the first user similarity between user n and user i; represents the value of the first user similarity between user i and user k; represents the user weight of the i-th user.

[0031] The first user feature vector and the user weight are concatenated to obtain a second user feature vector.

[0032] Based on the total information category and the second user feature vector, the users are classified to obtain the multiple user groups. The total information category refers to the total number of information categories obtained by classifying the information on the website. The user classification to obtain the multiple user groups includes:

[0033] Determine the total number of information categories. Based on the total number of information categories and the user weight, determine the second user similarity; the expression of the second user similarity is:

[0034] ;

[0035] in, Represents the second user similarity value between user n and user i; A second user feature vector representing user n; represents the second user feature vector of user i; j represents the commodity variable; C represents the total number of commodity categories; Feature weight coefficient representing information; Indicates taking the absolute value.

[0036] The users whose similarity to the second users is greater than a preset similarity threshold are divided into a user group. The preset similarity threshold refers to a preset maximum difference in similarity of the same user group.

[0037] Step 2: Based on the target user, determine the target user group. The target user refers to the user who needs to perform intelligent information push. The target user group may refer to the user group to which the target user belongs.

[0038] Step 3, obtaining the historical information browsing sequence and historical information recommendation sequence of the target user group. The historical information browsing sequence refers to the historical records of users browsing web pages on the website. Taking a shopping website as an example, the user's historical information browsing sequence may refer to the order in which the user browses products on the shopping website; further, a browsing time range may be set, and web pages within the browsing time range may be selected as web pages in the historical information browsing sequence. The historical information recommendation sequence refers to the web page browsing sequence recommended to the user by the website according to the user's preferences when the user browses web pages on the website.

[0039] Step 4, obtaining the current information browsing sequence of the target user, and generating multiple sets of current information recommendation sequences based on the current information browsing sequence; the information browsing sequence includes the website, browsing time and browsing action, etc. The current information browsing sequence refers to the sequence related to the order of browsing web pages obtained by the user currently browsing web pages on the website. The current information recommendation sequence refers to the sequence of web pages that the website currently recommends to the user. Taking a shopping website as an example, the products browsed by the user in sequence can be considered as the current information browsing sequence; the products recommended by the website to the user can be considered as the recommended current information. When the user clicks on different recommended information, the website will recommend subsequent products according to the user's preference. This continuous recommendation behavior can be considered as the current information recommendation sequence. The website address refers to the website address of the web page browsed by the user. The browsing time refers to the time the user browses the web page. The browsing action can refer to the operation performed by the user on the web page. Taking a shopping website as an example, the user's browsing actions include clicking, adding to cart, collecting, placing an order and sharing, etc. The generation of multiple sets of current information recommendation sequences based on the current information browsing sequence includes:

[0040] Based on the URL in the current information browsing sequence, the title and image of the website page are obtained.

[0041] Input the title into the word vector model to obtain the title vector sequence , represents the title vector variable; G represents the total number of web pages in the current information browsing sequence, Represents the title vector variable.

[0042] The image is input into the graph convolutional neural network to obtain an image vector sequence , Represents an image vector variable.

[0043] Concatenate the title vector sequence and the image vector sequence to obtain the current browsing vector sequence , where the current browsing vector variable .

[0044] Input the current browsing vector sequence into the long short-term memory network to obtain the hidden state of the current browsing sequence ,in, represents the gth hidden state in the current browsing vector, , Represents the g-1th hidden state in the current browsing vector.

[0045] Concatenate the hidden state of the current browsing sequence and the second user feature vector to obtain the current user feature vector ; represents the second user feature vector.

[0046] The current user feature vector and the information feature vector Matching is performed to obtain the plurality of sets of current information recommendation sequences. The information feature vector is obtained by learning the representation of the information knowledge graph, including:

[0047] The transE algorithm is used to perform representation learning on the information knowledge graph to obtain the entity representation matrix E and the relationship representation matrix R.

[0048] The information in the title is extracted to obtain the interest information of the target user. Taking a shopping website as an example, the title may refer to the product name of the product.

[0049] Based on the entity representation matrix and the relationship representation matrix of the interest information, an information feature vector related to the user's interest is established; the expression of the information feature vector is:

[0050] ;

[0051] in, represents the vth information feature vector; e represents the entity corresponding to the vth information, which refers to the information in the website; E represents the entity representation matrix; Indicate interest information; Shows the distance between information e and interest information in the information knowledge graph; Indicates the distance between the interest information and other information in the information knowledge graph; Indicates other information than information e; Represents the embedding of the vth information.

[0052] The obtaining of the multiple groups of current information recommendation sequences includes: using a multi-head attention mechanism to process the current user feature vector and the information feature vector to obtain multiple initial attention matrices. Using a cross-attention mechanism to process the multiple initial attention matrices to obtain a cross-attention matrix; inputting the cross-attention matrix into a fully connected layer to obtain a matching probability between the target user and the information; using the information with a matching probability greater than a preset matching threshold as the recommended information of the multiple groups of current information recommendation sequences; constructing a predicted browsing sequence based on the recommended information, using the predicted browsing sequence as a new current information browsing sequence for calculating new recommended information, and sorting them in the chronological order of obtaining the recommended information to obtain multiple groups of current information recommendation sequences. The preset matching threshold refers to the minimum value of the preset matching probability.

[0053] Step 5: Based on the historical information browsing sequence and the historical information recommendation sequence, the current information recommendation sequence is scored, and information is recommended to the target user according to the score; the score indicates the degree to which the current information recommendation sequence meets the information browsing preference of the target user. Information browsing preference refers to the user's preference for web page content. For example, what kind of product the user wants to buy.

[0054] The step of scoring the current information recommendation sequence includes:

[0055] Step 5.1, obtaining a second user similarity between the target user and other users in the target user group. Other users refer to users in the target user group other than the target user.

[0056] Step 5.2, determine the current group score of other users for the current information recommendation sequence and the group average score for the corresponding historical information recommendation sequence. The current group score can be used to indicate the preference of other users for the current information recommendation sequence. The current group score can be obtained through the browsing action of the user on the web page. Taking a shopping website as an example, the current group score can be determined based on whether the user browses the product corresponding to the web page in the current information recommendation sequence, whether the user purchases the product, whether the user orders the product, and whether the user's product browsing sequence conforms to or partially conforms to the current information recommendation sequence. The group average score refers to the average score of the user for each historical information browsing sequence. Taking a shopping website as an example, the user's click on the product can be used as the starting point of the information browsing sequence, and the last page of the website browsed before the user exits the website can be used as the end point of the information browsing sequence. By analyzing the user's actual browsing sequence and its conformity with the corresponding recommended browsing sequence, the user's group score for the historical information recommendation sequence is determined, and then the group score of the recommended sequence is averaged to obtain the group average score.

[0057] Step 5.3, determine the target average rating of the target user for the target historical information recommendation sequence. The target historical information recommendation sequence refers to the information browsing sequence recommended by the website to the target user when the target user browses the website. The target average rating refers to the average rating of the target user for the corresponding historical information recommendation sequence.

[0058] Step 5.4: Determine the score of the current information recommendation sequence based on the second user similarity, the current group score, the group average score and the target average score. The score of the current information recommendation sequence is expressed as:

[0059] ;

[0060] in, Indicates the score of the current information recommendation sequence; represents the target average rating; User variables representing the target user group; Represents the total number of users in the target user group; Indicates users in the target user group With target users The similarity of the second user; Indicates the difference between the current group score and the group average score.

[0061] Furthermore, the website is a shopping website; the information is commodities; the browsing actions include clicking, adding to cart and placing an order; and the commodity recommendation sequence is scored by analyzing the user's browsing actions on the recommended commodities.

[0062] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A website interface information intelligent recommendation system based on big data analysis, characterized in that: include: Cluster user portraits to obtain multiple user groups; Based on the target users, determine the target user group; Acquire the historical information browsing sequence and historical information recommendation sequence of the target user group; Acquire the current information browsing sequence of the target user, and generate multiple groups of current information recommendation sequences based on the current information browsing sequence; The information browsing sequence includes URL, browsing time and browsing actions; The generating a plurality of groups of current information recommendation sequences based on the current information browsing sequence comprises: Based on the URL in the current information browsing sequence, obtaining the title and image of the website page; Input the title into a word vector model to obtain a title vector sequence; Inputting the image into a graph convolutional neural network to obtain an image vector sequence; splicing the title vector sequence and the image vector sequence to obtain a current browsing vector sequence; Inputting the current browsing vector sequence into a long short-term memory network to obtain a hidden state of the current browsing sequence; Concatenate the current browsing sequence hidden state and the second user feature vector to obtain a current user feature vector; Matching the current user feature vector with the information feature vector to obtain the multiple groups of current information recommendation sequences; Based on the historical information browsing sequence and the historical information recommendation sequence, the current information recommendation sequence is scored, and information is recommended to the target user according to the score; the score indicates the degree to which the current information recommendation sequence meets the information browsing preference of the target user; wherein, scoring the current information recommendation sequence includes: Acquire a second user similarity between the target user and other users in the target user group; Determine the current group score of other users for the current information recommendation sequence and the group average score for the corresponding historical information recommendation sequence; Determine a target average score of the target user for the target historical information recommendation sequence; Based on the second user similarity, the current group score, the group average score and the target average score, determine the score of the current information recommendation sequence; the expression of the score of the current information recommendation sequence is: ； in, Indicates the score of the current information recommendation sequence; represents the target average rating; User variables representing the target user group; Represents the total number of users in the target user group; Indicates users in the target user group With target users The similarity of the second user; Indicates the difference between the current group score and the group average score.

2. The website interface information intelligent recommendation system based on big data analysis according to claim 1 is characterized in that: The user portraits are clustered to obtain multiple user groups, including: Based on the user portrait, construct a first user feature vector; Determining first user similarity by calculating the distance between the first user feature vectors; Determining a user weight based on the first user feature vector and the first user similarity; Concatenating the first user feature vector and the user weight to obtain a second user feature vector; Based on the total information category and the second user feature vector, the users are classified to obtain the multiple user groups.

3. The website interface information intelligent recommendation system based on big data analysis according to claim 2 is characterized in that: The expression of the first user similarity is: ； in, represents the value of the first user similarity between user n and user i, and Respectively represent different first user feature vectors, n and i represent user variables; m represents the first feature variable; M represents the total number of first features; Indicates the importance of the mth first feature; Represents the first feature and the first feature The distance between represents the mth first feature of the nth user; represents the mth first feature of the ith user.

4. The website interface information intelligent recommendation system based on big data analysis according to claim 2 is characterized in that: The expression for determining the user weight is: ； in, Indicates the user weight of the nth user; Indicates taking the absolute value; Represents the first user feature vector of the nth user The average distance to other users’ feature vectors; Represents the first user feature vector of the nth user The maximum distance to other user feature vectors; Represents the first user feature vector of the nth user The minimum value of the distance to other user feature vectors; m represents the first feature variable; M represents the total number of first features; Indicates the importance of the mth first feature; represents the mth first feature of the nth user; represents the mean of the mth first feature of all users; represents the maximum value of the mth first feature; represents the minimum value of the mth first feature; represents the L2 paradigm; represents the first user feature vector of the i-th user; represents the first user feature vector of the kth user; U represents the first user feature vector set ; Represents the first user feature vector of the i-th user The average distance to other users’ feature vectors; Represents the first user feature vector of the i-th user The maximum distance to other user feature vectors; Represents the first user feature vector of the i-th user The minimum distance to other user feature vectors; represents the value of the first user similarity between user n and user i; represents the value of the first user similarity between user i and user k; represents the user weight of the i-th user.

5. The website interface information intelligent recommendation system based on big data analysis according to claim 2 is characterized in that: The classifying of users to obtain the multiple user groups includes: Determine the total number of information categories; Based on the total number of the information categories and the user weight, a second user similarity is determined; the expression of the second user similarity is: ； in, Represents the second user similarity value between user n and user i; A second user feature vector representing user n; represents the second user feature vector of user i; j represents the commodity variable; C represents the total number of commodity categories; Feature weight coefficient representing information; Indicates taking the absolute value; The users whose second user similarity is greater than a preset similarity threshold are divided into a user group.

6. The website interface information intelligent recommendation system based on big data analysis according to claim 1 is characterized in that: The information feature vector is obtained by learning the information knowledge graph, including: Use the transE algorithm to perform representation learning on the information knowledge graph to obtain an entity representation matrix E and a relationship representation matrix R; Extracting information from the title to obtain interest information of the target user; Based on the entity representation matrix and the relationship representation matrix of the interest information, an information feature vector related to the user's interest is established; the expression of the information feature vector is: ； in, represents the vth information feature vector; e represents the entity corresponding to the vth information, which refers to the information in the website; E represents the entity representation matrix; Indicate interest information; Represents the distance between information e and interest information in the information knowledge graph; Indicates the distance between the interest information and other information in the information knowledge graph; Indicates other information than information e; Represents the embedding of the vth information.

7. The website interface information intelligent recommendation system based on big data analysis according to claim 1 is characterized in that: The obtaining of the plurality of sets of current information recommendation sequences comprises: Using a multi-head attention mechanism to process the current user feature vector and the information feature vector to obtain multiple initial attention matrices; Processing the multiple initial attention matrices using a cross attention mechanism to obtain a cross attention matrix; Input the cross attention matrix into a fully connected layer to obtain the matching probability between the target user and the information; The information with matching probability greater than a preset matching threshold is respectively used as the recommendation information of the multiple groups of current information recommendation sequences; Based on the recommendation information, a predicted browsing sequence is constructed, and the predicted browsing sequence is used as a new current information browsing sequence for calculating new recommendation information. The recommended information is sorted in the time sequence of obtaining the recommended information to obtain multiple groups of the current information recommendation sequences.

8. The website interface information intelligent recommendation system based on big data analysis according to claim 1 is characterized in that: The website is a shopping website; the information is commodities; the browsing actions include clicking, adding to cart and placing an order; and the commodity recommendation sequence is scored by analyzing the user's browsing actions on the recommended commodities.

Images