Information recommendation method and system based on artificial intelligence

Abstract

The invention provides an information recommendation method and system based on artificial intelligence, and the method comprises the steps: obtaining user preference data and user browsing data, the user preference data reveals a data type of user preference, the user browsing data comprises long-term browsing data and real-time browsing data, the long-term browsing data reveals a long-term browsing behavior of a user, and the real-time browsing data comprises a real-time browsing behavior of the user; the real-time browsing data reveals the real-time browsing behavior of the user; determining user characteristics of the user according to the user preference data and the user browsing data; matching the user characteristics of the user with the user characteristics of other users, and determining a target user; and determining to-be-recommended target information based on the target user, and recommending the target information to the user. Characteristic mining is carried out on long-term and real-time browsing data of the user through clustering analysis, the feature vectors are constructed in combination with user interests and preferences, the user interests can be considered, the cold start problem can be overcome as much as possible, and the information recommendation effect is effectively improved.

Description

Technical Field

[0001] The present application relates to the technical field of information recommendation, and in particular to an information recommendation method and system based on artificial intelligence. Background Art

[0002] With the development of the information age, traditional information recommendation solutions have gradually become unable to adapt to the load brought by the rapid increase in data volume, resulting in the original layout of information recommendation solutions becoming increasingly difficult to adapt to demand.

[0003] Traditional recommendation schemes include popularity-based recommendation schemes, similarity-based recommendation schemes, and user-based collaborative filtering recommendation schemes. Traditional recommendation schemes have indeed promoted the development of information recommendation services to a certain extent, but they each have their own limitations. For example, popularity-based recommendation schemes often only focus on the most popular content, but ignore the personalized needs of users; although similarity-based recommendation schemes take into account user historical preferences, they may fall into an "information cocoon" due to over-reliance on historical data, making it difficult to tap into users' potential interests. User-based collaborative filtering usually has a cold start problem, and the recommendation effect is poor when encountering new users due to the lack of sufficient historical data.

[0004] As a service platform, pushing information that is more in line with the reading interests of platform subscribers (such as novel documents, factual news, entertainment information, scientific exploration information, and other types of information in multiple fields) is an effective means to improve recommendation effects and increase user stickiness. Therefore, how to provide a more effective information recommendation solution is a problem that the service platform needs to solve. Summary of the invention

[0005] The purpose of the embodiments of the present application is to provide an information recommendation method and system based on artificial intelligence, so as to perform feature mining on users' long-term and real-time browsing data through cluster analysis, and at the same time combine users' interest preferences to jointly construct feature vectors as the basis of collaborative filtering analysis, which can not only take users' interests into consideration, but also overcome the cold start problem as much as possible, and effectively improve the information recommendation effect.

[0006] In order to achieve the above purpose, the embodiments of the present application are implemented in the following ways:

[0007] In a first aspect, an embodiment of the present application provides an artificial intelligence-based information recommendation method, including: obtaining user preference data and user browsing data, wherein the user preference data reveals the type of data preferred by the user, and the user browsing data includes long-term browsing data and real-time browsing data, the long-term browsing data is used to reveal the user's long-term browsing behavior, and the real-time browsing data is used to reveal the user's real-time browsing behavior; determining user characteristics of the user based on the user preference data and the user browsing data; matching the user characteristics of the user with the user characteristics of other users to determine the target user; determining target information to be recommended based on the target user, and recommending the target information to the user.

[0008] In combination with the first aspect, in a first possible implementation method of the first aspect, user characteristics of the user are determined based on user preference data and user browsing data, including: one-hot encoding the user preference data to obtain user preference characteristics; preprocessing and feature extraction of long-term browsing data in the user browsing data to determine long-term browsing characteristics; preprocessing and feature extraction of real-time browsing data in the user browsing data to determine real-time browsing characteristics; and combining user preference characteristics, long-term browsing characteristics, and real-time browsing characteristics into user characteristics.

[0009] In combination with the first possible implementation manner of the first aspect, in a second possible implementation manner of the first aspect, the long-term browsing data includes several pieces of long-term browsing information, and the long-term browsing data in the user browsing data is preprocessed and feature extracted to determine the long-term browsing features, including: preprocessing each piece of long-term browsing information in the long-term browsing data, wherein each piece of long-term browsing information corresponds to an information piece, and the long-term browsing information includes the time node, browsing duration, browsing entry, and operation frequency of the user browsing the corresponding information piece; for each preprocessed long-term browsing information: feature extraction is performed on the time node, browsing duration, browsing entry, and operation frequency in the long-term browsing information to determine the browsing feature component, and the browsing feature component is fused with the label feature component of the corresponding information information to obtain the long-term browsing vector corresponding to the long-term browsing information; the long-term browsing vectors corresponding to all the long-term browsing information are input into the first cluster analysis model, and the distance between each long-term browsing vector and each cluster center is output as the optimized long-term vector corresponding to each long-term browsing vector; based on each optimized long-term vector, the overall distribution feature of all optimized long-term vectors is extracted as the long-term browsing feature.

[0010] In combination with the second possible implementation manner of the first aspect, in a third possible implementation manner of the first aspect, based on each optimized long-term vector, the overall distribution characteristics of all optimized long-term vectors are extracted as long-term browsing features, including: for each dimension in the optimized long-term vector: respectively calculating the mean and standard deviation of all optimized long-term vectors in this dimension, and correspondingly obtaining the first eigenvalue and the second eigenvalue of this dimension; respectively determining the number of optimized long-term vectors with the minimum absolute value and the maximum absolute value in this dimension, and correspondingly obtaining the third eigenvalue and the fourth eigenvalue of this dimension; respectively determining the number of optimized long-term vectors with values ​​lower than the mean and lower than the set value in this dimension, and correspondingly obtaining the fifth eigenvalue and the sixth eigenvalue of this dimension; based on the first eigenvalue, the second eigenvalue, the third eigenvalue, the fourth eigenvalue, the fifth eigenvalue and the sixth eigenvalue in this dimension, determining the features in this dimension, and further integrating the features in all dimensions to obtain the long-term browsing features.

[0011] In combination with the third possible implementation method of the first aspect, in a fourth possible implementation method of the first aspect, based on the first eigenvalue, the second eigenvalue, the third eigenvalue, the fourth eigenvalue, the fifth eigenvalue and the sixth eigenvalue in the present dimension, the characteristics of the present dimension are determined, including: based on the sixth eigenvalue in each dimension, determining several target dimensions with the largest sixth eigenvalue value; for each target dimension: determining the target optimization long-term vector with a value lower than the set value in the current target dimension, and counting the mean and standard deviation of all target optimization long-term vectors in the current target dimension in each dimension as the supplementary eigenvalue in the corresponding dimension; for each dimension: based on the first eigenvalue, the second eigenvalue, the third eigenvalue, the fourth eigenvalue, the fifth eigenvalue, the sixth eigenvalue and all supplementary eigenvalues ​​in the present dimension, determining the characteristics of the present dimension.

[0012] In combination with the first possible implementation manner of the first aspect, in a fifth possible implementation manner of the first aspect, the real-time browsing data includes several pieces of real-time browsing information in a recent period of time, and the real-time browsing data in the user browsing data is preprocessed and feature extracted to determine the real-time browsing features, including: preprocessing each piece of real-time browsing information in the real-time browsing data, wherein each piece of real-time browsing information corresponds to a piece of information, and the real-time browsing information includes the time node of the user browsing the corresponding information information, the browsing time, the browsing entrance, the operation frequency, the bounce rate, the interactive behavior, the subscription status, the reading completion rate, the repeat visit rate, the page scroll depth, and the click-through conversion rate; for each piece of real-time browsing information after preprocessing: the time node in this piece of real-time browsing information , browsing time, browsing entry, operation frequency, bounce rate, interactive behavior, subscription status, reading completion rate, repeat visit rate, page scroll depth, and click-through conversion rate are used for feature extraction to determine the browsing feature components, and the browsing feature components are integrated with the title feature components, content feature components, tag feature components, and heat feature components of the corresponding information information to obtain the real-time browsing vector corresponding to this real-time browsing information; the real-time browsing vectors corresponding to all real-time browsing information are input into the second clustering analysis model, and the distance between each real-time browsing vector and each cluster center is output as the optimized real-time vector corresponding to each real-time browsing vector; based on each optimized real-time vector, the overall distribution characteristics of all optimized real-time vectors are extracted as the real-time browsing features.

[0013] In combination with the fifth possible implementation manner of the first aspect, in a sixth possible implementation manner of the first aspect, based on each optimized real-time vector, the overall distribution characteristics of all optimized real-time vectors are extracted as real-time browsing features, including: for each dimension in the optimized real-time vector: respectively calculating the mean and standard deviation of all optimized real-time vectors in this dimension, and correspondingly obtaining the first feature quantity and the second feature quantity of this dimension; respectively determining the number of optimized real-time vectors with the minimum absolute value and the maximum absolute value in this dimension, and correspondingly obtaining the third feature quantity and the fourth feature quantity of this dimension; respectively determining the number of optimized real-time vectors with values ​​lower than the mean and lower than the set value in this dimension, and correspondingly obtaining the fifth feature quantity and the sixth feature quantity of this dimension; based on the first feature quantity, the second feature quantity, the third feature quantity, the fourth feature quantity, the fifth feature quantity and the sixth feature quantity in this dimension, determining the features in this dimension, and further integrating the features in all dimensions to obtain the real-time browsing features.

[0014] In combination with the first possible implementation manner of the first aspect, in a seventh possible implementation manner of the first aspect, the user characteristics of the present user are matched with the user characteristics of other users to determine the target user, including: for each other user: calculating the similarity between the user preference characteristics of the present user and the user preference characteristics of the other user to obtain the preference similarity; calculating the similarity between the long-term browsing characteristics of the present user and the long-term browsing characteristics of the other user to obtain the long-term similarity; calculating the similarity between the real-time browsing characteristics of the present user and the real-time browsing characteristics of the other user to obtain the real-time similarity; based on the preference similarity, the long-term similarity and the real-time similarity, calculating the user similarity between the present user and the other user; and determining the target user based on the user similarity corresponding to each other user.

[0015] In combination with the first aspect, in an eighth possible implementation method of the first aspect, based on the target users, target information to be recommended is determined, including: obtaining a recommended list of information for each target user in a recent period of time; counting the number of recommendations, click-through rate, bounce rate, reading completion rate, repeat visit rate, and click-through conversion rate of each information; calculating the recommendation index of each information based on the number of recommendations, click-through rate, bounce rate, reading completion rate, repeat visit rate, and click-through conversion rate of each information; and determining the target information to be recommended based on the recommendation index of each information.

[0016] In a second aspect, an embodiment of the present application provides an artificial intelligence-based information recommendation system, comprising: a data acquisition module, used to acquire user preference data and user browsing data, wherein the user preference data reveals the type of data preferred by the user, and the user browsing data includes long-term browsing data and real-time browsing data, the long-term browsing data is used to reveal the user's long-term browsing behavior, and the real-time browsing data is used to reveal the user's real-time browsing behavior; a feature extraction module, used to determine the user characteristics of the user based on the user preference data and the user browsing data; a user matching module, used to match the user characteristics of the user with the user characteristics of other users to determine the target user; an information recommendation module, used to determine the target information to be recommended based on the target user, and recommend the target information to the user.

[0017] Beneficial effects:

[0018] (1) This solution obtains user preference data (determined based on the user's settings, access, sharing, and collection records, etc., which is generally generated by the platform based on the user's information. In this solution, the user preference data in the platform can be directly obtained) and user browsing data (including long-term browsing data and real-time browsing data. The long-term browsing data is used to reveal the user's long-term browsing behavior, and the real-time browsing data is used to reveal the user's real-time browsing behavior). Based on the user preference data and user browsing data, the user characteristics of the user are determined (the user preference characteristics are obtained by performing one-hot encoding on the user preference data; the long-term browsing data in the user browsing data is encoded by Preprocessing and feature extraction (including cluster analysis) to determine long-term browsing features; preprocessing and feature extraction (including cluster analysis) of real-time browsing data in user browsing data to determine real-time browsing features, which are finally combined into user features); matching the user features of this user with the user features of other users (performing corresponding similarity calculations) to determine the target user, and then based on the target user's information recommendation list (analyzing several indicators, such as the number of recommendations, click-through rate, bounce rate, reading completion rate, repeat visit rate, and click-through conversion rate of each information information), determine the target information to be recommended, and recommend the target information to this user. Through this solution, user preferences, long-term browsing data, and short-term browsing data (i.e., real-time browsing data) can be comprehensively analyzed, and feature mining of users' long-term and real-time browsing data can be performed through cluster analysis. At the same time, combined with user interest preferences, feature vectors are jointly constructed as the basis for collaborative filtering analysis. This method can not only take user interests into consideration, but also overcome the cold start problem as much as possible, effectively improving the information recommendation effect.

[0019] (2) In the feature extraction of long-term browsing data, each long-term browsing information (including the time node when the user browsed the corresponding information information, browsing time, browsing entry and operation frequency) is pre-processed, and then preliminary feature extraction is performed (for example, unified format and standardization processing are performed to obtain the corresponding browsing feature component, and then it is merged with the label feature component of the corresponding information information) to obtain the long-term browsing vector corresponding to this long-term browsing information, and the long-term browsing vectors corresponding to all long-term browsing information are input into the first clustering analysis model (the clustering analysis model of this embodiment is the K-Means clustering algorithm, which is obtained through training and regularly optimized through the long-term browsing data of all users in the platform), and cluster analysis is performed to output the distance between each long-term browsing vector and each cluster center (a multi-dimensional vector will be formed, and the dimension corresponds to the number of cluster centers in the first clustering analysis model) as the optimized long-term vector corresponding to each long-term browsing vector, and further extract the overall distribution characteristics of all optimized long-term vectors as the long-term browsing feature. Specifically, the mean and standard deviation of all optimized long-term vectors in this dimension are calculated, and the first and second eigenvalues ​​of this dimension are obtained; the number of optimized long-term vectors with the minimum absolute value and the maximum absolute value in this dimension are determined, and the third and fourth eigenvalues ​​of this dimension are obtained; the number of optimized long-term vectors with values ​​lower than the mean and lower than the set value in this dimension are determined, and the fifth and sixth eigenvalues ​​of this dimension are obtained; based on the first eigenvalue, the second eigenvalue, the third eigenvalue, the fourth eigenvalue, the fifth eigenvalue and the sixth eigenvalue in this dimension, the features in this dimension are determined, and the features in all dimensions are further integrated to obtain the long-term browsing features. In this way, not only can the first clustering analysis model be used to perform clustering analysis on the large number of long-term browsing information of this user with different quantities, and further mine potential features, but also the distribution statistics of the optimized long-term vectors corresponding to the output of each long-term browsing information can be performed (reflected in the form of quantity or percentage, and normalization can be performed when the quantity is used for statistics), so as to overcome the cold start problem caused by the small amount of data as much as possible.When constructing long-term browsing features, by analyzing the data of some key dimensions (key cluster centers, which can reflect the user's personalization to a certain extent), for example, based on the sixth eigenvalue on each dimension (that is, the number of optimized long-term vectors that meet the set value on a certain dimension, which can reflect the concentration, that is, a large number of long-term browsing information is gathered in the cluster center), several target dimensions with the largest sixth eigenvalue values ​​(for example, the top three) are determined; for each target dimension: the target optimized long-term vector with a value lower than the set value on the current target dimension is determined, and the mean and standard deviation of all target optimized long-term vectors on the current target dimension on each dimension are counted as the supplementary eigenvalues ​​on the corresponding dimension; for each dimension: based on the first eigenvalue, the second eigenvalue, the third eigenvalue, the fourth eigenvalue, the fifth eigenvalue, the sixth eigenvalue and all supplementary eigenvalues ​​on this dimension, the features on this dimension are determined. In this way, not only the overall distribution of the user's long-term browsing data is considered, but also the potential personalized features of the user are considered, which can further enhance the reflection ability of the long-term browsing features, so as to more accurately match similar users for users (do collaborative filtering based on users) and improve the information recommendation effect.

[0020] (3) In the feature extraction of real-time browsing data, each real-time browsing information in the real-time browsing data is preprocessed. Each real-time browsing information corresponds to a piece of news information. The real-time browsing information includes the time node, browsing duration, browsing entry, operation frequency, bounce rate, interactive behavior, subscription status, reading completion rate, repeat visit rate, page scroll depth, and click-through conversion rate of the user browsing the corresponding news information. Feature extraction is performed to determine the browsing feature component, and the browsing feature component is fused with the title feature component, content feature component, label feature component, and heat feature component of the corresponding news information to obtain the real-time browsing vector corresponding to the real-time browsing information. The real-time browsing information contains more detailed indicators (compared to the long-term browsing information), which can well reflect the user's real-time reading and operation status, and the corresponding news information contains richer feature information (including title feature component, content feature component, label feature component, heat feature component, etc., which can be obtained by obtaining the relevant data of the corresponding news information through the platform), so as to better reflect the user's current status and interests. Afterwards, the real-time browsing vectors corresponding to all the real-time browsing information are input into the second cluster analysis model (obtained by training using the K-Means clustering algorithm), and the distance between each real-time browsing vector and each cluster center is output as the optimized real-time vector corresponding to each real-time browsing vector; based on each optimized real-time vector, the overall distribution characteristics of all optimized real-time vectors are extracted as real-time browsing features. For each dimension in the optimized real-time vector: the mean and standard deviation of all optimized real-time vectors in this dimension are calculated respectively, and the first feature quantity and the second feature quantity of this dimension are obtained; the number of optimized real-time vectors with the minimum absolute value and the maximum absolute value in this dimension are determined respectively, and the third feature quantity and the fourth feature quantity of this dimension are obtained respectively; the number of optimized real-time vectors with values ​​lower than the mean and lower than the set value in this dimension are determined respectively, and the fifth feature quantity and the sixth feature quantity of this dimension are obtained respectively; based on the first feature quantity, the second feature quantity, the third feature quantity, the fourth feature quantity, the fifth feature quantity and the sixth feature quantity in this dimension, the features in this dimension are determined, and the features in all dimensions are further integrated to obtain the real-time browsing features. In this way, not only can the second clustering analysis model be used to perform cluster analysis on different quantities of real-time browsing information of the user to further explore potential features, but also distribution statistics can be performed based on the optimized real-time vectors output corresponding to each real-time browsing information (which can also be reflected in the form of quantity or percentage, and normalization can be performed when using quantity for statistics), so as to overcome the cold start problem caused by the small amount of data as much as possible.

[0021] (4) When performing user matching and information recommendation, the preference similarity between the user preference characteristics of the user and the user preference characteristics of other users is calculated; the long-term similarity between the long-term browsing characteristics of the user and the long-term browsing characteristics of other users is calculated; the real-time similarity between the real-time browsing characteristics of the user and the real-time browsing characteristics of other users is calculated; based on the preference similarity, long-term similarity and real-time similarity, the user similarity between the user and the other user is calculated; based on the user similarity corresponding to each other user, the target user is determined. In this way, multiple characteristics can be combined to determine the target user who is more matched in all aspects, which is very suitable for platforms with an increasing number of users. When performing information recommendation, the recommended list of information information for each target user in the recent period is obtained; the number of recommended people, click-through rate, bounce rate, reading completion rate, repeat visit rate, and click-through conversion rate of each information information are counted, and the recommendation index of each information information is calculated to determine the target information to be recommended. In this way, information information can be better recommended to users and the recommendation effect can be effectively improved.

[0022] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0024] Figure 1 A flowchart of an information recommendation method based on artificial intelligence provided in an embodiment of the present application.

[0025] Figure 2 A schematic diagram for obtaining access-related data of some users.

[0026] Figure 3 Schematic diagram of subscribing to related data for some users.

[0027] Figure 4 This is a structural block diagram of an information recommendation system based on artificial intelligence.

[0028] Icons: 10-information recommendation system based on artificial intelligence; 11-data acquisition module; 12-feature extraction module; 13-user matching module; 14-information recommendation module. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0030] In order to recommend information to platform users, this embodiment provides an information recommendation method based on artificial intelligence. Figure 1 , Figure 1 Flow chart of the information recommendation method based on artificial intelligence. In this embodiment, the information recommendation method based on artificial intelligence may include step S10, step S20, step S30, and step S40.

[0031] First, when a user browses news information, relevant news information may be recommended to the user. At this time, step S10 may be executed.

[0032] Step S10: Obtain user preference data and user browsing data, wherein the user preference data reveals the type of data preferred by the user, and the user browsing data includes long-term browsing data and real-time browsing data, the long-term browsing data is used to reveal the user's long-term browsing behavior, and the real-time browsing data is used to reveal the user's real-time browsing behavior.

[0033] In this embodiment, user preference data and user browsing data may be acquired.

[0034] The user preference data reveals the type of data that the user prefers, which can be determined through the user's settings, access, sharing, subscription, collection records, etc. Generally, the platform generates a user profile based on the user's information and determines the corresponding user preference data based on the user profile. In this embodiment, the platform can directly obtain the user's user preference data, which will not be described in detail.

[0035] User browsing data includes long-term browsing data and real-time browsing data. Long-term browsing data is used to reveal users' long-term browsing behavior, and real-time browsing data is used to reveal users' real-time browsing behavior.

[0036] Long-term browsing data may include several long-term browsing information, each of which corresponds to one piece of information. The long-term browsing information includes the time node, browsing duration, browsing entry and operation frequency when the user browses the corresponding information. The information information corresponding to the long-term browsing information, each piece of information may include a corresponding label feature component (each piece of information on the platform actually has many features, such as title features, several content features, label features, popularity features, etc., but to analyze long-term browsing information, so many features are not needed, only the label feature component needs to be obtained).

[0037] Real-time browsing data can include real-time browsing information within a recent time period (e.g., the last ten minutes, the last half hour, etc. Of course, other solutions can also fix the number of browsed information, such as the last 10, but this solution easily ignores time continuity and has certain analytical defects. To overcome this, the continuity of these browsing information in the time dimension must also be considered. Therefore, this solution uses time periods as an example for explanation). Each real-time browsing information corresponds to an information message. The real-time browsing information includes the time node when the user browses the corresponding information message, the browsing duration, the browsing entry (i.e., how to enter the browsing, such as search entry, history entry, recommendation entry, etc.), the operation frequency (the number of clicks, slides, scrolls, etc. on the page by the user, reflecting the user's activity and participation), the bounce rate (the proportion of users who exit after browsing only one page, reflecting the attractiveness of the page content to the user), interactive behaviors (such as likes, comments, shares, reports, etc.), subscription status (reflecting the user's subscription information. This platform is a variety of subscription packages launched, which can largely reflect the user's The information content includes the following: the degree of liking of information), reading completion rate (the proportion of users who read the entire article or watch the entire video or listen to the entire music, which measures the attractiveness of the information content and the reading depth and immersion depth of the users), repeat visit rate (the proportion of users who visit the same information multiple times, which reflects the users' interest or confusion in the content), page scrolling depth (the proportion or distance of users scrolling the page, or the length of time users watch videos or listen to music, which reflects the users' attention to the content and browsing depth), click-through conversion rate (the proportion of users who click on specific buttons, such as reading the full text, sharing, commenting, etc., which measures the attractiveness of information and user participation), etc. To analyze real-time browsing information, it is necessary to consider the characteristics of the information, so it is necessary to obtain as much information as possible, such as title feature components, content feature components, tag feature components, and heat feature components.

[0038] like Figure 2 and Figure 3 As shown, Figure 2 and Figure 3 A schematic diagram of some user access and subscription-related data is shown (the interface is derived from the subscription data management subsystem independently developed by the applicant, which is used to provide some data support for the platform). Many data are difficult to display directly through the display interface of the platform system and are not suitable for display. Therefore, only a small number of examples are given, and this part of the data has been protected for user privacy and does not expose user privacy information.

[0039] After obtaining the user preference data and the user browsing data, step S20 may be executed.

[0040] Step S20: Determine the user characteristics of the user according to the user preference data and the user browsing data.

[0041] In this embodiment, the user characteristics of the user are determined based on the user preference data and the user browsing data.

[0042] First, the user preference data can be one-hot encoded to obtain the user preference features. This part of the technology is very simple, which is equivalent to providing a fully labeled data. If the user has a certain label, the value is determined to be 1 in the corresponding dimension, otherwise it is 0. The obtained feature vector is the user preference feature.

[0043] Secondly, the long-term browsing data in the user browsing data can be preprocessed and feature extracted to determine the long-term browsing features.

[0044] Exemplarily, each long-term browsing information in the long-term browsing data may be preprocessed, where each long-term browsing information corresponds to one piece of information, and the long-term browsing information includes the time node, browsing duration, browsing entry, and operation frequency when the user browses the corresponding information.

[0045] Then, for each long-term browsing information after preprocessing: feature extraction can be performed on the time node, browsing duration, browsing entry and operation frequency in this long-term browsing information, the browsing feature component can be determined, and the browsing feature component can be fused with the label feature component of the corresponding information information to obtain the long-term browsing vector corresponding to this long-term browsing information.

[0046] Specifically, the time nodes in each long-term browsing information can be processed in a unified manner, and the browsing time nodes can be converted into a unified format (such as UNIX timestamp), and time-related features can be extracted as needed, such as time periods in a day (morning, afternoon, evening), weekdays and weekends, months, seasons, etc., to capture the browsing preferences of users in different time periods; and the browsing duration can be counted (accurate to seconds), the browsing entrances can be assigned values ​​(different values ​​are assigned according to different browsing entrances, for example, the search entrance is assigned a value of 1, the history entrance is assigned a value of 2, the interface recommendation entrance is assigned a value of 3, the associated recommendation entrance is assigned a value of 4, etc.) and the operation frequency (the number of operations per second or per minute is counted). These initially extracted browsing feature components are fused (for example, spliced) with the label feature components of the corresponding information information to obtain the long-term browsing vector corresponding to each long-term browsing information. Each long-term browsing vector obtained after such processing is in a unified format for subsequent clustering analysis.

[0047] After obtaining the long-term browsing vector corresponding to each long-term browsing information, the long-term browsing vector corresponding to all the long-term browsing information of the user can be input into the first cluster analysis model (using the K-Means clustering algorithm, obtained through training, and regularly tuned through the long-term browsing data of all users in the platform, for example, updated once a month). Then, through cluster analysis, the distance between each long-term browsing vector and each cluster center in the first cluster analysis model can be calculated, thereby obtaining a vector as the optimized long-term vector corresponding to the long-term browsing vector. Assuming that the number of cluster centers of the first cluster analysis model is k and the number of long-term browsing vectors is n, n optimized long-term vectors can be obtained.

[0048] Afterwards, based on each optimized long-term vector, the overall distribution characteristics of all optimized long-term vectors can be extracted as long-term browsing features.

[0049] Exemplarily, for optimizing each dimension in the long-term vector:

[0050] The mean and standard deviation of all optimized long-term vectors in this dimension can be calculated respectively, and the first eigenvalue and the second eigenvalue of this dimension can be obtained accordingly. For example, the i-th optimized long-term vector is recorded as r i =(x i1 ,x i2 ,…,x ij ,…,x ik ), i∈[1,n], then, for the jth dimension, j∈[1,k], the following formula is used to calculate the mean and standard deviation of all optimized long-term vectors of this user in this dimension j:

[0051]

[0052] Among them, μ j is the mean of all optimized long-term vectors of this user on dimension j, n is the number of long-term browsing vectors of this user, x ij For the value of the i-th optimized long-term vector in dimension j, σ j is the standard deviation of all optimized long-term vectors of this user in dimension j.

[0053] And, the number of optimized long-term vectors with the minimum absolute value and the maximum absolute value in this dimension can be determined respectively, and the third eigenvalue and the fourth eigenvalue of this dimension can be obtained accordingly. For example, the dimension where the minimum absolute value and the maximum absolute value of each optimized long-term vector (the elements of the optimized long-term vector in this embodiment are all non-negative numbers, so the absolute value is itself) are located is determined, and then for each dimension (taking the jth dimension as an example), the number of optimized long-term vectors with the minimum absolute value in this dimension is counted. j (You can also calculate the proportion to weaken the difference in quantity, for example ) and the number of optimized long-term vectors b with the largest absolute value in this dimension j (You can also calculate the proportion to weaken the difference in quantity, for example ), as the third eigenvalue (accounting for α j as the seventh eigenvalue) and the fourth eigenvalue (accounting for β j as the eighth eigenvalue).

[0054] The number of optimized long-term vectors with values ​​lower than the mean and lower than the set value (for example, the mean calculated by the minimum absolute value element of each optimized long-term vector) in this dimension can also be determined respectively, and the fifth eigenvalue and the sixth eigenvalue of this dimension are obtained accordingly. For example, for each dimension (taking the jth dimension as an example), the number of optimized long-term vectors with values ​​lower than the mean μ in this dimension is counted. j The number of optimized term vectors c j (You can also calculate the proportion to weaken the difference in quantity, for example ), as the fifth eigenvalue (accounting for γ j As the ninth eigenvalue), count the number of optimized long-term vectors d that are lower than the set value in this dimension j (You can also calculate the proportion to weaken the difference in quantity, for example ), as the sixth eigenvalue (accounting for δ j as the tenth eigenvalue).

[0055] Afterwards, the features on this dimension can be determined based on the first eigenvalue, the second eigenvalue, the third eigenvalue, the fourth eigenvalue, the fifth eigenvalue and the sixth eigenvalue on this dimension, and the features on all dimensions can be further integrated to obtain the long-term browsing features.

[0056] Exemplarily, based on the sixth eigenvalue of each dimension, several target dimensions (eg, 1, 3, 5, etc., 3 being taken as an example in this embodiment) having the largest sixth eigenvalue value can be determined.

[0057] Then, for each target dimension: the target optimization long-term vector (for example, there are y target optimization long-term vectors) with a value lower than the set value (the set value of the current target dimension) in the current target dimension can be determined, and the mean and standard deviation of all target optimization long-term vectors (that is, these y target optimization long-term vectors) in each dimension of the current target dimension are counted as supplementary eigenvalues ​​in the corresponding dimension.

[0058] Finally, for each dimension: based on the first eigenvalue, the second eigenvalue, the third eigenvalue (and the seventh eigenvalue), the fourth eigenvalue (and the eighth eigenvalue), the fifth eigenvalue (and the ninth eigenvalue), the sixth eigenvalue (and the tenth eigenvalue) and all supplementary eigenvalues ​​on the dimension, the feature of the dimension can be determined. Finally, a long-term browsing feature containing a total of k dimensions and each dimension containing z elements can be obtained.

[0059] After determining the long-term browsing characteristics of the user, the real-time browsing data in the user's browsing data may be preprocessed and feature extracted to determine the real-time browsing characteristics.

[0060] Exemplarily, each real-time browsing information in the real-time browsing data can be preprocessed. Here, each real-time browsing information corresponds to a piece of information, and the real-time browsing information includes the time node when the user browses the corresponding information, browsing time, browsing entry, operation frequency, bounce rate, interactive behavior, subscription status, reading completion rate, repeat visit rate, page scroll depth, and click-through conversion rate.

[0061] Then, for each real-time browsing information after preprocessing: extract features of the time node, browsing time, browsing entry, operation frequency, bounce rate, interactive behavior, subscription status, reading completion rate, repeat visit rate, page scroll depth, and click-through conversion rate in this real-time browsing information, determine the browsing feature component, and merge the browsing feature component with the title feature component, content feature component, tag feature component, and popularity feature component of the corresponding news information to obtain the real-time browsing vector corresponding to this real-time browsing information.

[0062] Similarly, the time nodes in each real-time browsing information can be processed in a unified manner, converted into a unified format (such as UNIX timestamp), and time-related features can be extracted as needed, such as time periods of the day (morning, afternoon, evening), weekdays and weekends, months, seasons, etc.; in addition, the browsing time can be counted (accurate to seconds), the browsing entrances can be assigned values ​​(different values ​​are assigned according to different browsing entrances, for example, the search entrance is assigned a value of 1, the history entrance is assigned a value of 2, the interface recommendation entrance is assigned a value of 3, the associated recommendation entrance is assigned a value of 4, etc.) and the operation frequency (counting the number of operations per second or per minute). Also, assign values ​​to the bounce rate (since statistics are collected here for each real-time browsing information, a value of 1 is assigned to a page that bounces after browsing, and a value of 0 is assigned to other cases), interactive behavior (a value of 0 is assigned to no interactive behavior, and a value is assigned to different interactive behaviors if there are interactive behaviors. If there are multiple interactive behaviors, a value is assigned based on the existing behaviors. For example, if both values ​​1 and 2 exist, it is recorded as 12), subscription status (a value of 1 is assigned to a subscription, and a value of 0 is assigned to no subscription), reading completion rate (a value of 1 is assigned to a completed reading, and a value of 0 is assigned to an incomplete reading), and repeat visit rate (whether it is a repeated visit). The feature extraction is performed based on the information information of the page, if it is, it is recorded as 1, otherwise it is recorded as 0), page scrolling depth (value is assigned according to the ratio of scrolling depth to total depth), click conversion rate (if no relevant button is clicked, it is recorded as 0, if relevant button is clicked, it is assigned according to the different button types clicked, and if there are multiple click behaviors, it is assigned according to the combination of existing click behaviors, for example, if there are both 1 and 2 assigned, it is recorded as 12) and then fused (for example, spliced) with the title feature component, content feature component, tag feature component, and heat feature component of the corresponding information information to obtain the real-time browsing vector corresponding to this real-time browsing information. Each real-time browsing vector obtained after this processing is in a unified format for subsequent clustering analysis.

[0063] Based on this, the real-time browsing vectors corresponding to all real-time browsing information can be input into the second clustering analysis model (which also uses the K-Means clustering algorithm and is obtained through training, but the training data is different from the first clustering analysis model), and the distance between each real-time browsing vector and each cluster center is output as the optimized real-time vector corresponding to each real-time browsing vector.

[0064] Exemplarily, for each dimension in the optimized real-time vector: the mean and standard deviation of all optimized real-time vectors in this dimension can be calculated respectively, and the first characteristic value and the second characteristic value of this dimension are obtained accordingly. The number of optimized real-time vectors with the minimum absolute value and the maximum absolute value in this dimension are determined respectively, and the third characteristic value and the fourth characteristic value of this dimension are obtained accordingly. The number of optimized real-time vectors with values ​​lower than the mean and lower than the set value in this dimension are determined respectively, and the fifth characteristic value and the sixth characteristic value of this dimension are obtained accordingly. For the specific calculation and determination methods, please refer to the processing process of optimizing long-term vectors in the previous text, which will not be repeated here.

[0065] After that, the features of this dimension can be determined based on the first feature quantity, the second feature quantity, the third feature quantity, the fourth feature quantity, the fifth feature quantity and the sixth feature quantity of this dimension, and the features of all dimensions can be further integrated to obtain the real-time browsing features. The difference between this and the long-term browsing features mentioned above is that since the optimization of the real-time vector itself takes into account many real-time detailed features (such as user operations, related features of information, etc.), there is no need to calculate and integrate the supplementary feature values.

[0066] After obtaining the real-time browsing features of the user, the user preference features, long-term browsing features and real-time browsing features can be combined into user features (no need to splice because the user preference features, long-term browsing features and real-time browsing features differ in dimensions or number of attributes).

[0067] After obtaining the user characteristics, step S30 may be executed.

[0068] Step S30: Match the user characteristics of the current user with the user characteristics of other users to determine the target user.

[0069] In this embodiment, the user characteristics of the current user may be matched with the user characteristics of other users. Then, for each other user:

[0070] First, the similarity between the user preference feature of the user and the user preference feature of the other user may be calculated (for example, by calculating the cosine similarity) to obtain the preference similarity.

[0071] Secondly, the similarity between the long-term browsing characteristics of the current user and the long-term browsing characteristics of other users can be calculated: specifically, the similarity score on each dimension can be measured by cosine similarity, and then the similarity score of each dimension can be combined and weighted summed (the weight can be measured according to the set value of each dimension, and the weight can be allocated according to the proportion of the set value of each dimension, which can further strengthen the personalized characteristics of the user so as to match more similar users), and the comprehensive similarity score can be obtained as the long-term similarity. For example,

[0072] Furthermore, the similarity between the real-time browsing features of the user and the real-time browsing features of other users can be calculated: Similarly, the similarity score on each dimension can be measured by cosine similarity, and then the similarity score of each dimension can be combined to perform weighted summation (the weight can be measured according to the set value of each dimension, and the weight is allocated according to the proportion of the set value of each dimension, which can further enhance the user's personalized characteristics so as to match more similar users), and the comprehensive similarity score is obtained as the real-time similarity.

[0073] Finally, based on preference similarity, long-term similarity and real-time similarity, the user similarity between the user and the other user is calculated. Here, the overall similarity is obtained by weighted summation. As user similarity, the weight design can be pre-set, for example, preference similarity accounts for 0.3, long-term similarity accounts for 0.3, and real-time similarity accounts for 0.4. It is also possible to design dynamically adjusted weights, dynamically assign weights according to the importance of each similarity or a certain indicator, for example, preference similarity accounts for 0.3, and the remaining 0.7 is dynamically assigned according to the size of preference similarity, and corresponding weights are assigned to long-term similarity and real-time similarity, which is not limited here.

[0074] Afterwards, the target user may be determined based on the user similarity corresponding to each other user (for example, 100 other users with the highest user similarity may be determined as the target user).

[0075] After the target user is determined, step S40 may be executed.

[0076] Step S40: Based on the target user, determine the target information to be recommended, and recommend the target information to the user.

[0077] In this embodiment, a list of recommended information for each target user in a recent period of time (e.g., the last hour, or the 100 most recently viewed information) can be obtained. Then, the number of recommended people for each information (i.e., the number of people each information is recommended to these target users), click-through rate (the ratio of users clicking on this information to the number of recommended people for this information), bounce rate (the probability of browsing no more than one page or exiting within a certain period of time after each information is recommended to the target user and clicked), reading completion rate (the probability of browsing completion after each information is recommended to the target user and clicked), repeat visit rate (the probability of information being recommended to the user for the first time and clicked by the user), and click-through conversion rate (the probability of further clicks such as viewing the full text, continuing to watch, etc. after each information is recommended to the target user and clicked).

[0078] By determining the number of recommended persons, click-through rate, bounce rate, reading completion rate, repeat visit rate, and click-through conversion rate of each information information, the recommendation index of each information information can be comprehensively calculated. The calculation method of this recommendation index is related to the specific type of information information (such as text information, video, music, real-time news, popular science documents, etc.). Different types of information information have different calculation methods. The measurement method of these recommendation indexes is not the focus of the present invention, and can also be flexibly changed under the influence of various factors such as different needs and application scenarios, which will not be elaborated here.

[0079] After determining the recommendation index of each piece of information, several pieces of information with the highest recommendation index (eg, 5, 10, 30) may be determined as target information to be recommended and recommended to the user.

[0080] See also Figure 4 Based on the same inventive concept, this embodiment also provides an artificial intelligence-based information recommendation system 10, including:

[0081] The data acquisition module 11 is used to acquire user preference data and user browsing data, wherein the user preference data reveals the type of data preferred by the user, and the user browsing data includes long-term browsing data and real-time browsing data. The long-term browsing data is used to reveal the user's long-term browsing behavior, and the real-time browsing data is used to reveal the user's real-time browsing behavior.

[0082] The feature extraction module 12 is used to determine the user features of the user according to the user preference data and the user browsing data.

[0083] The user matching module 13 is used to match the user characteristics of the current user with the user characteristics of other users to determine the target user.

[0084] The information recommendation module 14 is used to determine target information to be recommended based on the target user and recommend the target information to the user.

[0085] In this embodiment, the feature extraction module 12 is specifically used to: perform one-hot encoding on the user preference data to obtain the user preference features; perform preprocessing and feature extraction on the long-term browsing data in the user browsing data to determine the long-term browsing features; perform preprocessing and feature extraction on the real-time browsing data in the user browsing data to determine the real-time browsing features; and combine the user preference features, the long-term browsing features and the real-time browsing features into user features.

[0086] In this embodiment, the feature extraction module 12, the long-term browsing data includes a plurality of long-term browsing information, which is specifically used to: pre-process each long-term browsing information in the long-term browsing data, wherein each long-term browsing information corresponds to an information information, and the long-term browsing information includes the time node, browsing duration, browsing entry and operation frequency of the user browsing the corresponding information information; for each pre-processed long-term browsing information: extract features of the time node, browsing duration, browsing entry and operation frequency in this long-term browsing information, determine the browsing feature component, and fuse the browsing feature component with the label feature component of the corresponding information information to obtain the long-term browsing vector corresponding to this long-term browsing information; input the long-term browsing vectors corresponding to all the long-term browsing information into the first cluster analysis model, output the distance between each long-term browsing vector and each cluster center as the optimized long-term vector corresponding to each long-term browsing vector; based on each optimized long-term vector, extract the overall distribution characteristics of all optimized long-term vectors as the long-term browsing feature.

[0087] In this embodiment, the feature extraction module 12 is specifically used to: for each dimension in the optimized long-term vector: calculate the mean and standard deviation of all optimized long-term vectors in this dimension respectively, and obtain the first eigenvalue and the second eigenvalue of this dimension accordingly; determine the number of optimized long-term vectors with the minimum absolute value and the maximum absolute value in this dimension respectively, and obtain the third eigenvalue and the fourth eigenvalue of this dimension accordingly; determine the number of optimized long-term vectors with values ​​lower than the mean and lower than the set value in this dimension respectively, and obtain the fifth eigenvalue and the sixth eigenvalue of this dimension accordingly; based on the first eigenvalue, the second eigenvalue, the third eigenvalue, the fourth eigenvalue, the fifth eigenvalue and the sixth eigenvalue in this dimension, determine the features in this dimension, and further integrate the features in all dimensions to obtain the long-term browsing features.

[0088] In this embodiment, the feature extraction module 12 is specifically used to: determine several target dimensions with the largest sixth eigenvalue based on the sixth eigenvalue on each dimension; for each target dimension: determine the target optimization long-term vector with a value lower than the set value on the current target dimension, and count the mean and standard deviation of all target optimization long-term vectors on the current target dimension on each dimension as the supplementary eigenvalue on the corresponding dimension; for each dimension: determine the features on this dimension based on the first eigenvalue, second eigenvalue, third eigenvalue, fourth eigenvalue, fifth eigenvalue, sixth eigenvalue and all supplementary eigenvalues ​​on this dimension.

[0089] In this embodiment, the real-time browsing data includes a number of real-time browsing information in a recent period of time, and the feature extraction module 12 is specifically used to: pre-process each real-time browsing information in the real-time browsing data, wherein each real-time browsing information corresponds to an information information, and the real-time browsing information includes the time node, browsing time, browsing entrance, operation frequency, bounce rate, interactive behavior, subscription status, reading completion rate, repeat visit rate, page scroll depth, and click-through conversion rate of the user browsing the corresponding information information; for each real-time browsing information after pre-processing: the time node, browsing time, browsing entrance, operation frequency, bounce rate, interactive behavior, subscription status, and page scroll depth in this real-time browsing information are pre-processed. The real-time browsing information is extracted based on the situation, reading completion rate, repeat visit rate, page scrolling depth, and click-through conversion rate to determine the browsing feature components, and the browsing feature components are integrated with the title feature components, content feature components, tag feature components, and heat feature components of the corresponding news information to obtain the real-time browsing vector corresponding to the real-time browsing information; the real-time browsing vectors corresponding to all the real-time browsing information are input into the second clustering analysis model, and the distance between each real-time browsing vector and each cluster center is output as the optimized real-time vector corresponding to each real-time browsing vector; based on each optimized real-time vector, the overall distribution characteristics of all optimized real-time vectors are extracted as the real-time browsing features.

[0090] In this embodiment, the feature extraction module 12 is specifically used to: for each dimension in the optimized real-time vector: calculate the mean and standard deviation of all optimized real-time vectors in this dimension respectively, and obtain the first feature value and the second feature value of this dimension accordingly; determine the number of optimized real-time vectors with the minimum absolute value and the maximum absolute value in this dimension respectively, and obtain the third feature value and the fourth feature value of this dimension accordingly; determine the number of optimized real-time vectors with values ​​lower than the mean and the set value in this dimension respectively, and obtain the fifth feature value and the sixth feature value of this dimension accordingly; based on the first feature value, the second feature value, the third feature value, the fourth feature value, the fifth feature value and the sixth feature value in this dimension, determine the features in this dimension, and further integrate the features in all dimensions to obtain the real-time browsing features.

[0091] In this embodiment, the user matching module 13 is specifically used to: for each other user: calculate the similarity between the user preference characteristics of the present user and the user preference characteristics of this other user to obtain the preference similarity; calculate the similarity between the long-term browsing characteristics of the present user and the long-term browsing characteristics of this other user to obtain the long-term similarity; calculate the similarity between the real-time browsing characteristics of the present user and the real-time browsing characteristics of this other user to obtain the real-time similarity; based on the preference similarity, the long-term similarity and the real-time similarity, calculate the user similarity between the present user and the other user; based on the user similarity corresponding to each other user, determine the target user.

[0092] In this embodiment, the information recommendation module 14 is specifically used to: obtain a list of recommended information for each target user in a recent period of time; count the number of recommendations, click-through rate, bounce rate, reading completion rate, repeat visit rate, and click-through conversion rate of each information; calculate the recommendation index of each information based on the number of recommendations, click-through rate, bounce rate, reading completion rate, repeat visit rate, and click-through conversion rate of each information; determine the target information to be recommended based on the recommendation index of each information.

[0093] In summary, the embodiment of the present application provides an information recommendation method and system based on artificial intelligence, by obtaining user preference data (determined according to the user's settings, access, sharing, collection records, etc., generally generated by the platform according to the user's information, and in this solution, the user preference data in the platform can be directly obtained) and user browsing data (including long-term browsing data and real-time browsing data, the long-term browsing data is used to reveal the user's long-term browsing behavior, and the real-time browsing data is used to reveal the user's real-time browsing behavior), according to the user preference data and the user browsing data, the user characteristics of the user are determined (the user preference characteristics are obtained by performing unique hot encoding on the user preference data; the user browsing data is used to perform unique hot encoding on the user preference data, and ... Preprocess and extract features (including cluster analysis) of long-term browsing data in browsing data to determine long-term browsing features; preprocess and extract features (including cluster analysis) of real-time browsing data in user browsing data to determine real-time browsing features, and finally combine them into user features); match the user features of this user with the user features of other users (calculate the corresponding similarity) to determine the target user, and then determine the target information to be recommended based on the target user's information recommendation list (analyze several indicators, such as the number of recommended people, click-through rate, bounce rate, reading completion rate, repeat visit rate, and click-through conversion rate of each information information), and recommend the target information to this user. Through this solution, user preferences, long-term browsing data, and short-term browsing data (i.e., real-time browsing data) can be comprehensively analyzed, and feature mining of users' long-term and real-time browsing data can be performed through cluster analysis. At the same time, combined with user interest preferences, feature vectors are jointly constructed as the basis for collaborative filtering analysis. This method can not only take user interests into consideration, but also overcome the cold start problem as much as possible, and effectively improve the information recommendation effect.

[0094] In the feature extraction of long-term browsing data, each long-term browsing information (including the time node when the user browsed the corresponding information information, browsing time, browsing entry and operation frequency) is preprocessed, and then preliminary feature extraction is performed (for example, unified format and standardization processing are performed to obtain the corresponding browsing feature components, and then merged with the label feature components of the corresponding information information) to obtain the long-term browsing vector corresponding to this long-term browsing information, and the long-term browsing vectors corresponding to all long-term browsing information are input into the first clustering analysis model (the clustering analysis model of this embodiment is the K-Means clustering algorithm, which is obtained through training and regularly tuned through the long-term browsing data of all users in the platform), and cluster analysis is performed to output the distance between each long-term browsing vector and each cluster center (a multidimensional vector will be formed, and the dimension corresponds to the number of cluster centers in the first clustering analysis model) as the optimized long-term vector corresponding to each long-term browsing vector, and further extract the overall distribution characteristics of all optimized long-term vectors as long-term browsing features. Specifically, the mean and standard deviation of all optimized long-term vectors in this dimension are calculated, and the first and second eigenvalues ​​of this dimension are obtained; the number of optimized long-term vectors with the minimum absolute value and the maximum absolute value in this dimension are determined, and the third and fourth eigenvalues ​​of this dimension are obtained; the number of optimized long-term vectors with values ​​lower than the mean and lower than the set value in this dimension are determined, and the fifth and sixth eigenvalues ​​of this dimension are obtained; based on the first eigenvalue, the second eigenvalue, the third eigenvalue, the fourth eigenvalue, the fifth eigenvalue and the sixth eigenvalue in this dimension, the features in this dimension are determined, and the features in all dimensions are further integrated to obtain the long-term browsing features. In this way, not only can the first clustering analysis model be used to perform clustering analysis on the large number of long-term browsing information of this user with different quantities, and further mine potential features, but also the distribution statistics of the optimized long-term vectors corresponding to the output of each long-term browsing information can be performed (reflected in the form of quantity or percentage, and normalization can be performed when the quantity is used for statistics), so as to overcome the cold start problem caused by the small amount of data as much as possible.When constructing long-term browsing features, by analyzing the data of some key dimensions (key cluster centers, which can reflect the user's personalization to a certain extent), for example, based on the sixth eigenvalue on each dimension (that is, the number of optimized long-term vectors that meet the set value on a certain dimension, which can reflect the concentration, that is, a large number of long-term browsing information is gathered in the cluster center), several target dimensions with the largest sixth eigenvalue values ​​(for example, the top three) are determined; for each target dimension: the target optimized long-term vector with a value lower than the set value on the current target dimension is determined, and the mean and standard deviation of all target optimized long-term vectors on the current target dimension on each dimension are counted as the supplementary eigenvalues ​​on the corresponding dimension; for each dimension: based on the first eigenvalue, the second eigenvalue, the third eigenvalue, the fourth eigenvalue, the fifth eigenvalue, the sixth eigenvalue and all supplementary eigenvalues ​​on this dimension, the features on this dimension are determined. In this way, not only the overall distribution of the user's long-term browsing data is considered, but also the potential personalized features of the user are considered, which can further enhance the reflection ability of the long-term browsing features, so as to more accurately match similar users for users (do collaborative filtering based on users) and improve the information recommendation effect.

[0095] In the feature extraction of real-time browsing data, each real-time browsing information in the real-time browsing data is preprocessed. Each real-time browsing information corresponds to an information information. The real-time browsing information includes the time node, browsing duration, browsing entrance, operation frequency, bounce rate, interactive behavior, subscription status, reading completion rate, repeat visit rate, page scroll depth, and click-through conversion rate of the user browsing the corresponding information information. Feature extraction is performed to determine the browsing feature component, and the browsing feature component is fused with the title feature component, content feature component, label feature component, and heat feature component of the corresponding information information to obtain the real-time browsing vector corresponding to this real-time browsing information. The real-time browsing information contains more detailed indicators (compared with long-term browsing information), which can well reflect the user's real-time reading and operation situation, and the corresponding information information contains richer feature information (including title feature component, content feature component, label feature component, heat feature component, etc., which can be obtained by obtaining relevant data of the corresponding information information through the platform), so as to better reflect the user's current status and interests. Afterwards, the real-time browsing vectors corresponding to all the real-time browsing information are input into the second cluster analysis model (obtained by training using the K-Means clustering algorithm), and the distance between each real-time browsing vector and each cluster center is output as the optimized real-time vector corresponding to each real-time browsing vector; based on each optimized real-time vector, the overall distribution characteristics of all optimized real-time vectors are extracted as real-time browsing features. For each dimension in the optimized real-time vector: the mean and standard deviation of all optimized real-time vectors in this dimension are calculated respectively, and the first feature quantity and the second feature quantity of this dimension are obtained; the number of optimized real-time vectors with the minimum absolute value and the maximum absolute value in this dimension are determined respectively, and the third feature quantity and the fourth feature quantity of this dimension are obtained respectively; the number of optimized real-time vectors with values ​​lower than the mean and lower than the set value in this dimension are determined respectively, and the fifth feature quantity and the sixth feature quantity of this dimension are obtained respectively; based on the first feature quantity, the second feature quantity, the third feature quantity, the fourth feature quantity, the fifth feature quantity and the sixth feature quantity in this dimension, the features in this dimension are determined, and the features in all dimensions are further integrated to obtain the real-time browsing features. In this way, not only can the second clustering analysis model be used to perform cluster analysis on different quantities of real-time browsing information of the user to further explore potential features, but also distribution statistics can be performed based on the optimized real-time vectors output corresponding to each real-time browsing information (which can also be reflected in the form of quantity or percentage, and normalization can be performed when using quantity for statistics), so as to overcome the cold start problem caused by the small amount of data as much as possible.

[0096] When performing user matching and information recommendation, the preference similarity between the user's user preference characteristics and other users' user preference characteristics is calculated; the long-term similarity between the user's long-term browsing characteristics and other users' long-term browsing characteristics is calculated; the real-time similarity between the user's real-time browsing characteristics and other users' real-time browsing characteristics is calculated; based on the preference similarity, long-term similarity and real-time similarity, the user similarity between the user and other users is calculated; based on the user similarity corresponding to each other user, the target user is determined. In this way, multiple features can be combined to determine the target user who is more matched in all aspects, which is very suitable for platforms with an increasing number of users. When recommending information, the recommended list of information information for each target user in the recent period is obtained; the number of recommended people, click-through rate, bounce rate, reading completion rate, repeat visit rate, and click-through conversion rate of each information information are counted, and the recommendation index of each information information is calculated to determine the target information to be recommended. In this way, information information can be better recommended to users and the recommendation effect can be effectively improved.

[0097] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0098] The above description is only an embodiment of the present application and is not intended to limit the protection scope of the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An information recommendation method based on artificial intelligence, characterized in that: include: Obtain user preference data and user browsing data, wherein the user preference data reveals the type of data preferred by the user, and the user browsing data includes long-term browsing data and real-time browsing data, the long-term browsing data is used to reveal the user's long-term browsing behavior, and the real-time browsing data is used to reveal the user's real-time browsing behavior; Determine the user characteristics of the user based on the user preference data and the user browsing data; Match the user characteristics of the current user with the user characteristics of other users to determine the target user; Based on the target user, the target information to be recommended is determined, and the target information is recommended to the user.

2. The information recommendation method based on artificial intelligence according to claim 1, characterized in that: Based on the user preference data and user browsing data, the user characteristics of the user are determined, including: Perform one-hot encoding on user preference data to obtain user preference features; Preprocess and extract features of long-term browsing data in user browsing data to determine long-term browsing features; Preprocessing and feature extraction of real-time browsing data in user browsing data to determine real-time browsing features; Based on user preference characteristics, long-term browsing characteristics and real-time browsing characteristics, they are combined into user characteristics.

3. The information recommendation method based on artificial intelligence according to claim 2, characterized in that: The long-term browsing data includes several pieces of long-term browsing information. The long-term browsing data in the user browsing data is preprocessed and feature extracted to determine the long-term browsing features, including: Preprocessing each piece of long-term browsing information in the long-term browsing data, wherein each piece of long-term browsing information corresponds to one piece of news information, and the long-term browsing information includes the time node, browsing duration, browsing entry, and operation frequency of the user browsing the corresponding news information; For each long-term browsing information after preprocessing: extract features of the time node, browsing duration, browsing entry and operation frequency in this long-term browsing information, determine the browsing feature component, and merge the browsing feature component with the label feature component of the corresponding news information to obtain the long-term browsing vector corresponding to this long-term browsing information; Inputting the long-term browsing vectors corresponding to all the long-term browsing information into the first cluster analysis model, and outputting the distance between each long-term browsing vector and each cluster center as the optimized long-term vector corresponding to each long-term browsing vector; Based on each optimized long-term vector, the overall distribution characteristics of all optimized long-term vectors are extracted as long-term browsing features.

4. The information recommendation method based on artificial intelligence according to claim 3, characterized in that: Based on each optimized long-term vector, the overall distribution characteristics of all optimized long-term vectors are extracted as long-term browsing features, including: For each dimension in the long-term vector, optimize: Calculate the mean and standard deviation of all optimized long-term vectors in this dimension respectively, and obtain the first eigenvalue and the second eigenvalue of this dimension accordingly; The number of optimized long-term vectors with the minimum absolute value and the maximum absolute value in this dimension are determined respectively, and the third eigenvalue and the fourth eigenvalue of this dimension are obtained correspondingly; The number of optimized long-term vectors with values ​​lower than the mean and lower than the set value in this dimension is determined respectively, and the fifth eigenvalue and the sixth eigenvalue of this dimension are obtained correspondingly; Based on the first eigenvalue, the second eigenvalue, the third eigenvalue, the fourth eigenvalue, the fifth eigenvalue and the sixth eigenvalue of this dimension, the feature of this dimension is determined, and the features of all dimensions are further integrated to obtain the long-term browsing feature.

5. The information recommendation method based on artificial intelligence according to claim 4, characterized in that: Based on the first eigenvalue, the second eigenvalue, the third eigenvalue, the fourth eigenvalue, the fifth eigenvalue and the sixth eigenvalue in this dimension, the features in this dimension are determined, including: Based on the sixth eigenvalue of each dimension, determining several target dimensions with the largest sixth eigenvalue values; For each target dimension: determine the target optimization long-term vector with a value lower than the set value in the current target dimension, and count the mean and standard deviation of all target optimization long-term vectors in each dimension in the current target dimension as the supplementary eigenvalues ​​in the corresponding dimension; For each dimension: based on the first eigenvalue, the second eigenvalue, the third eigenvalue, the fourth eigenvalue, the fifth eigenvalue, the sixth eigenvalue and all supplementary eigenvalues ​​on the dimension, determine the characteristics of the dimension.

6. The information recommendation method based on artificial intelligence according to claim 2, characterized in that: The real-time browsing data includes a number of real-time browsing information in the recent period. The real-time browsing data in the user browsing data is preprocessed and feature extracted to determine the real-time browsing features, including: Preprocess each piece of real-time browsing information in the real-time browsing data, wherein each piece of real-time browsing information corresponds to one piece of news information, and the real-time browsing information includes the time node when the user browses the corresponding news information, browsing duration, browsing entry, operation frequency, bounce rate, interactive behavior, subscription status, reading completion rate, repeat visit rate, page scroll depth, and click-through conversion rate; For each real-time browsing information after preprocessing: extract features of the time node, browsing time, browsing entry, operation frequency, bounce rate, interactive behavior, subscription status, reading completion rate, repeat visit rate, page scroll depth, and click-through conversion rate in this real-time browsing information, determine the browsing feature component, and fuse the browsing feature component with the title feature component, content feature component, tag feature component, and popularity feature component of the corresponding news information to obtain the real-time browsing vector corresponding to this real-time browsing information; Inputting the real-time browsing vectors corresponding to all the real-time browsing information into the second cluster analysis model, and outputting the distance between each real-time browsing vector and each cluster center as the optimized real-time vector corresponding to each real-time browsing vector; Based on each optimized real-time vector, the overall distribution characteristics of all optimized real-time vectors are extracted as real-time browsing features.

7. The information recommendation method based on artificial intelligence according to claim 6, characterized in that: Based on each optimized real-time vector, the overall distribution features of all optimized real-time vectors are extracted as real-time browsing features, including: For each dimension in the optimized real-time vector: Calculate the mean and standard deviation of all optimized real-time vectors in this dimension respectively, and obtain the first feature quantity and the second feature quantity of this dimension accordingly; The number of optimized real-time vectors having the minimum absolute value and the maximum absolute value in the dimension is determined respectively, and the third feature quantity and the fourth feature quantity in the dimension are obtained correspondingly; The number of optimized real-time vectors having values ​​lower than the mean and lower than the set value in the dimension is determined respectively, and the fifth characteristic value and the sixth characteristic value of the dimension are obtained accordingly; Based on the first feature quantity, the second feature quantity, the third feature quantity, the fourth feature quantity, the fifth feature quantity and the sixth feature quantity in this dimension, the feature in this dimension is determined, and the features in all dimensions are further integrated to obtain the real-time browsing feature.

8. The information recommendation method based on artificial intelligence according to claim 2, characterized in that: Match the user characteristics of this user with those of other users to identify the target user, including: For each additional user: Calculate the similarity between the user preference feature of the user and the user preference feature of the other user to obtain the preference similarity; Calculate the similarity between the long-term browsing characteristics of the user and the long-term browsing characteristics of other users to obtain the long-term similarity; Calculate the similarity between the real-time browsing characteristics of the user and the real-time browsing characteristics of the other user to obtain the real-time similarity; Based on preference similarity, long-term similarity and real-time similarity, calculate the user similarity between the user and the other user; Based on the user similarity corresponding to each other user, the target user is determined.

9. The information recommendation method based on artificial intelligence according to claim 1, characterized in that: Based on the target user, determine the target information to be recommended, including: Obtain a list of recommended news information for each target user in the recent period; Statistics on the number of referrals, click-through rate, bounce rate, reading completion rate, repeat visit rate, and click-through conversion rate for each piece of information; Calculate the recommendation index of each piece of information based on the number of recommendations, click-through rate, bounce rate, reading completion rate, repeat visit rate, and click-through conversion rate of each piece of information; According to the recommendation index of each information, the target information to be recommended is determined.

10. An information recommendation system based on artificial intelligence, characterized in that: include: A data acquisition module is used to acquire user preference data and user browsing data, wherein the user preference data reveals the type of data preferred by the user, and the user browsing data includes long-term browsing data and real-time browsing data, the long-term browsing data is used to reveal the user's long-term browsing behavior, and the real-time browsing data is used to reveal the user's real-time browsing behavior; A feature extraction module is used to determine the user features of the user based on the user preference data and the user browsing data; The user matching module is used to match the user characteristics of the current user with the user characteristics of other users to determine the target user; The information recommendation module is used to determine the target information to be recommended based on the target user and recommend the target information to the user.

Images