User active relationship discovery method and system for information recommendation
By constructing a multi-attribute heterogeneous graph of user relationships and a logistic regression model, user relationship scores are calculated, and positive relationship groups are identified. This solves the problem that existing information recommendation systems cannot effectively utilize user relationships, and improves the accuracy and timeliness of information recommendations.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ARMY MEDICAL UNIV
- Filing Date
- 2023-01-06
- Publication Date
- 2026-05-12
AI Technical Summary
Existing information recommendation systems cannot effectively utilize user relationships, especially positive relationships, resulting in low recommendation accuracy and an inability to capture dynamic changes in user needs.
By constructing a multi-attribute heterogeneous graph of user relationships, the relationship score between the target user and other users is calculated. Based on the score threshold and the TOP-N algorithm, the positive relationship groups of the target user are determined, and information is recommended.
It enables the dynamic capture of user needs based on changes in user relationships, thereby improving the accuracy and timeliness of information recommendations.
Smart Images
Figure CN115982469B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of information recommendation technology, and in particular to a method and system for discovering positive user relationships for information recommendation. Background Technology
[0002] With the deepening research into large-scale data management, recommender systems have been recognized as the most effective method for solving the data overload problem. In recent years, the recommendation accuracy of recommender systems has improved rapidly. However, most current methods focus on static user-information matrix analysis, paying insufficient attention to the dynamic changes in user interests, which significantly impacts the performance of information recommendation systems.
[0003] In addition, user relationships are playing an increasingly important role in information recommendation, especially positive relationships (such as friendships). However, traditional information recommendation methods cannot effectively utilize user relationships for recommendations, and they are not sensitive to dynamic changes in user preferences and user relationships, failing to capture dynamic changes in user needs. This results in low accuracy of information recommendations, which in turn has an adverse impact on information utilization.
[0004] Therefore, how to more accurately discover positive user relationships to help improve the accuracy of information recommendation is a problem that urgently needs to be solved in this field. Summary of the Invention
[0005] This invention aims to solve the technical problems existing in the prior art, and innovatively proposes a method and system for discovering positive user relationships for information recommendation. By calculating the positive relationship groups of target users based on the association relationships between users, the positive user relationships are discovered, thereby helping to improve the accuracy of information recommendation.
[0006] To achieve the above-mentioned objectives of the present invention, according to a first aspect of the present invention, the present invention provides a method for discovering positive user relationships for information recommendation, the method comprising the following steps:
[0007] Construct a multi-attribute heterogeneous graph of user relationships, wherein the multi-attribute heterogeneous graph of user relationships uses users and information as vertices and the lines connecting the vertices as edges, the edges are used to represent the relationships between the vertices, the relationships include the relationships between users and the relationships between users and information;
[0008] Calculate the relationship score between the target user and other users based on the relationships between vertices in the aforementioned multi-attribute heterogeneous graph of user relationships.
[0009] Based on preset scoring thresholds and the score of the target user's relationship with other users, a positive relationship group for the target user is obtained;
[0010] Based on the TOP-N algorithm, N users most similar to the target user are determined from the target user's active relationship group, and these N users are selected as active relationship users for information recommendation to the target user, where N is a positive integer.
[0011] Preferably, the construction of the multi-attribute heterogeneous graph of user relationships includes:
[0012] A heterogeneous graph of user relationships is constructed by using users and information as vertices and the lines connecting the vertices as edges.
[0013] Based on the user's attribute characteristics, edge attributes and vertex attributes are added to the heterogeneous graph of user relationships to obtain the multi-attribute heterogeneous graph of user relationships. The user's attribute characteristics include the number of groups the user participates in, the number of common groups among users, the size of common groups among users, and the number of common tags among users. The tags refer to the labels that users assign to information.
[0014] Preferably, the step of calculating the relationship score between the target user and other users based on the relationships between vertices in the multi-attribute heterogeneous graph of user relationships includes:
[0015] Based on the logistic regression model, the user relationship score calculation model is obtained by fusing features from the user's attribute characteristics, including the number of groups the user participates in, the number of common groups among users, the size of common groups among users, and the number of common tags among users.
[0016] The relationship score between the target user and other users is calculated based on the user relationship score calculation model and the user attribute characteristics of the target user.
[0017] Preferably, the user relationship score calculation model is as follows:
[0018]
[0019] Among them, score(v i ,v j ) represents the score of the relationship between user i and user j, |Gr(v i | represents the number of groups that user i participates in, |Gr(v j )| represents the number of groups that user j participates in, Cg(v i ,v j ) represents the number of common groups between user i and user j, Cgs(v i ,v j ) k Ct(v) represents the size of the common group between user i and user j. i ,vj ) represents the number of common labels between user i and user j, and n represents the number of common groups between user i and user j, i.e., n = Cg(v i ,v j ).
[0020] Preferably, obtaining the target user's positive relationship group based on a preset scoring threshold and the relationship score between the target user and other users includes:
[0021] The relationship score between the target user and other users is compared with a preset score threshold, and users whose relationship score with other users exceeds the preset score threshold are grouped into the same group to form the target user's positive relationship group.
[0022] According to a second aspect of the present invention, the present invention also provides a user-positive relationship discovery system for information recommendation, the system comprising:
[0023] The heterogeneous graph construction module is used to construct a multi-attribute heterogeneous graph of user relationships, wherein the multi-attribute heterogeneous graph of user relationships uses users and information as vertices and the lines connecting the vertices as edges. The edges are used to represent the relationships between the vertices, including the relationships between users and the relationships between users and information.
[0024] The user relationship calculation module is used to calculate the relationship score between the target user and other users based on the relationship between vertices in the multi-attribute heterogeneous graph of user relationships.
[0025] The positive relationship group generation module is used to generate positive relationship groups for a target user based on a preset scoring threshold and the score of the relationship between the target user and other users.
[0026] The active relationship user identification module determines the N users most similar to the target user from the active relationship group of the target user based on the TOP-N algorithm, and uses the N users most similar to the target user as active relationship users for information recommendation to the target user, where N is a positive integer.
[0027] Preferably, the heterogeneous graph construction module is specifically used to:
[0028] A heterogeneous graph of user relationships is constructed by using users and information as vertices and the lines connecting the vertices as edges.
[0029] Based on the user's attribute characteristics, edge attributes and vertex attributes are added to the heterogeneous graph of user relationships to obtain the multi-attribute heterogeneous graph of user relationships. The user's attribute characteristics include the number of groups the user participates in, the number of common groups among users, the size of common groups among users, and the number of common tags among users. The tags refer to the labels that users assign to information.
[0030] Preferably, when the user relationship calculation module performs the calculation of the relationship score between the target user and other users based on the relationship between vertices in the multi-attribute heterogeneous graph of user relationships, the specific user is:
[0031] Based on the logistic regression model, the user relationship score calculation model is obtained by fusing features from the user's attribute characteristics, including the number of groups the user participates in, the number of common groups among users, the size of common groups among users, and the number of common tags among users.
[0032] The relationship score between the target user and other users is calculated based on the user relationship score calculation model and the user attribute characteristics of the target user.
[0033] Preferably, the user relationship score calculation model is as follows:
[0034]
[0035] Among them, score(v i ,v j ) represents the score of the relationship between user i and user j, |Gr(v i | represents the number of groups that user i participates in, |Gr(v j )| represents the number of groups that user j participates in, Cg(v i ,v j ) represents the number of common groups between user i and user j, Cgs(v i ,v j ) k Ct(v) represents the size of the common group between user i and user j. i ,v j ) represents the number of common labels between user i and user j, and n represents the number of common groups between user i and user j, i.e., n = Cg(v i ,v j ).
[0036] Preferably, the positive relationship group generation module, when performing the operation of obtaining the positive relationship group of the target user based on a preset scoring threshold and the relationship score between the target user and other users, specifically uses the following methods:
[0037] The relationship score between the target user and other users is compared with a preset score threshold, and users whose relationship score with other users exceeds the preset score threshold are grouped into the same group to form the target user's positive relationship group.
[0038] As can be seen from the above scheme, the present invention provides a method and system for discovering positive user relationships for information recommendation. It constructs a multi-attribute heterogeneous graph of user relationships, where users and information are vertices, and the lines connecting the vertices are edges representing the relationships between the vertices. These relationships include relationships between users and relationships between users and information. Based on the relationships between vertices in the multi-attribute heterogeneous graph, a score is calculated for the relationship between the target user and other users. Based on a preset score threshold and the relationship scores between the target user and other users, a positive relationship group for the target user is obtained. Using the TOP-N algorithm, the N users most similar to the target user are determined from the positive relationship group, and these N users are selected as the positive relationship users for information recommendation to the target user. The present invention analyzes the relationships between users by constructing a multi-attribute heterogeneous graph of user relationships, effectively responding to dynamic changes in user interests and user relationships, thereby effectively capturing dynamic changes in user needs. By calculating the positive relationship group of the target user based on the relationships between users, the present invention discovers positive user relationships and assists in timely and accurate information recommendation.
[0039] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description
[0040] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:
[0041] Figure 1 This is a flowchart of a preferred embodiment of the present invention for a method of discovering positive user relationships for information recommendation;
[0042] Figure 2 This is a schematic diagram of a preferred embodiment of the present invention for a user-positive relationship discovery system for information recommendation. Detailed Implementation
[0043] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
[0044] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the meaning consistent with their meaning in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined.
[0045] like Figure 1 The diagram shown is a flowchart of a preferred embodiment of a user-positive relationship discovery method for information recommendation according to the present invention. The method may include the following steps:
[0046] S101, Construct a multi-attribute heterogeneous graph of user relationships, wherein the multi-attribute heterogeneous graph of user relationships uses users and information as vertices and the lines connecting the vertices as edges, the edges are used to represent the relationships between the vertices, the relationships include the relationships between users and the relationships between users and information;
[0047] With the rapid development of online social networks, people are more inclined to trust recommendations from users with positive relationships. Discovering user relationships, especially positive relationships, can effectively improve the accuracy of information recommendations. To better discover positive user relationships and thus assist in timely and accurate information recommendations, it is first necessary to construct a multi-attribute heterogeneous graph of user relationships. Heterogeneous graphs contain various types of nodes and relationships. Compared to homogeneous graphs containing only one type of node / relationship, heterogeneous graphs are more complex and contain more information. Therefore, this embodiment of the invention uses a multi-attribute heterogeneous graph of user relationships to analyze the connections between users, which can effectively respond to the dynamic changes in user interests and user relationships, thereby effectively capturing the dynamic changes in user needs. Specifically, in this multi-attribute heterogeneous graph of user relationships, users and information are used as vertices, and the lines connecting the vertices are used as edges. The edges represent the relationships between vertices, including relationships between users and relationships between users and information.
[0048] S102, Calculate the relationship score between the target user and other users based on the relationship between vertices in the user relationship multi-attribute heterogeneous graph;
[0049] After constructing the multi-attribute heterogeneous graph of user relationships, the relationships between users are analyzed using this graph. Since the relationship score between a target user and other users can accurately and objectively reflect the positiveness of the relationship, this embodiment calculates the relationship score between the target user and other users based on the relationships between vertices in the multi-attribute heterogeneous graph, thereby quantifying the positiveness of the relationship between the target user and other users.
[0050] S103, Based on the preset scoring threshold and the score of the relationship between the target user and other users, obtain the positive relationship group of the target user;
[0051] Since the relationship score between the target user and other users can accurately and objectively reflect the positiveness of the relationship between the target user and other users, after calculating the relationship score between the target user and other users, it is necessary to obtain the target user's positive relationship group based on the preset score threshold and the relationship score between the target user and other users. That is, to judge other users with positive relationships with the target user based on the relationship score between the target user and other users, the method of setting a score threshold is used to judge other users with positive relationships with the target user, and all users judged to have positive relationships with the target user are grouped together to obtain the target user's positive relationship group.
[0052] S104, Based on the TOP-N algorithm, determine the N users most similar to the target user from the active relationship group of the target user, and use the N users most similar to the target user as active relationship users for information recommendation to the target user, where N is a positive integer.
[0053] After obtaining the target user's active relationship group, to reduce the computational load of information recommendation and achieve faster and more accurate recommendations, it is necessary to determine the N users most similar to the target user from the target user's active relationship group based on the TOP-N algorithm. This involves sorting the users in the target user's active relationship group from highest to lowest relationship score and selecting the top N users as the target user's most similar users. These N most similar users are then used as active relationship users for information recommendations to the target user. This further filters the users in the target user's active relationship group using the TOP-N algorithm, narrowing down the range of other users with positive relationships to the target user for information recommendations. It should be noted that N is specifically set as needed.
[0054] In summary, this embodiment provides a method for discovering positive user relationships for information recommendation. First, a multi-attribute heterogeneous graph of user relationships is constructed, where users and information are vertices, and the lines connecting the vertices are edges representing the relationships between the vertices. These relationships include those between users and those between users and information. Then, based on the relationships between the vertices in the multi-attribute heterogeneous graph, a score is calculated indicating the relationship between the target user and other users. Next, based on a preset score threshold and the score indicating the relationship between the target user and other users, a positive relationship group for the target user is obtained. Finally, based on the TOP-N algorithm, the N users most similar to the target user are determined from the positive relationship group, and these N users are selected as the positive relationship users for information recommendation to the target user. This embodiment analyzes the relationships between users by constructing a multi-attribute heterogeneous graph of user relationships, effectively responding to dynamic changes in user interests and user relationships, thereby effectively capturing dynamic changes in user needs. By calculating the positive relationship group of the target user based on the relationships between users, it discovers positive user relationships and assists in timely and accurate information recommendation.
[0055] In other embodiments of the present invention, based on the above embodiments, the construction of a multi-attribute heterogeneous graph of user relationships includes:
[0056] A heterogeneous graph of user relationships is constructed by using users and information as vertices and the lines connecting the vertices as edges.
[0057] Based on the user's attribute characteristics, edge attributes and vertex attributes are added to the heterogeneous graph of user relationships to obtain the multi-attribute heterogeneous graph of user relationships. The user's attribute characteristics include the number of groups the user participates in, the number of common groups among users, the size of common groups among users, and the number of common tags among users. The tags refer to the labels that users assign to information.
[0058] Specifically, in this embodiment, the user relationship multi-attribute heterogeneous graph G = (V, E, Gr, T) is constructed, where V represents the vertex set, E represents the edge set, Gr represents the group set, and T represents the set of labels that users have made to the information.
[0059] Since the existence of user associations depends on users' activities on the network, the determination of the existence of user associations can be linked to the discrimination of many user characteristics (i.e., user attribute features). However, no single user attribute is sufficient to predict the existence of user associations. This embodiment of the invention uses the following user attribute features to describe user activities:
[0060] (1) Number of groups participated in by users
[0061] The more groups a user participates in, the more active that user is in the network, and therefore the more likely that user is to connect with other users and have more relationships. Specifically, in this embodiment, the number of groups that user i participates in is represented as |Gr(v i )|, where Gr(v i Let represent the set of groups that user i participates in, and | represent the value operation, which yields the number of groups that user i participates in. Specifically, the group refers to the community to which the user belongs in the network, calculated by the community detection algorithm.
[0062] (2) Number of public groups among users
[0063] Since a higher number of common groups among users indicates greater similarity, the number of common groups is an important feature for predicting positive relationships. Specifically, in this embodiment, the number of common groups between user i and user j is represented as Cg(v i ,v j ),in,
[0064] Cg(v i ,v j )=|Gr(v i )∩Gr(v j )|
[0065] Among them, Gr(v i Gr(v) represents the set of groups that user i participates in. j Gr(v) represents the set of groups that user j participates in, ∩ represents the intersection operation, and Gr(v) represents the set of groups that user j participates in. i )∩Gr(v j ) represents the set of groups that users i and j both participate in, and | represents the value operation, which gives the number of groups that users i and j both participate in.
[0066] (3) Size of public groups among users
[0067] If two users join a small public group, the likelihood of a positive relationship between them is high. This is because a large public group is likely a user group for a popular product or a large organization where users have loosely coupled relationships. Specifically, the group size characterizes the number of users in the group; the more users in a group, the larger the group. In this embodiment, the public group size between user i and user j is represented as Cgs(v i ,v j ),in,
[0068] Cgs(v i ,v j =Sizeof(Gr(v i )∩Gr(v j )).
[0069] (4) Number of public tags among users
[0070] Two users in the same group who share many common tags (common tags refer to the same or similar labels applied to the same information by both users; for example, if two users simultaneously give the same or similar reviews of the same product, then this same or similar review is a common tag shared by the two users) are more similar. Therefore, more common tags mean that the two users have more similar interests and are more likely to have a positive relationship. Specifically, the number of common tags between user i and user j is denoted as Ct(v i ,v j ),in,
[0071] Ct(v i ,v j )=|Ct(v i )∩Ct(v j )|
[0072] Wherein, Ct(v i Ct(v) represents the set of tags that user i has assigned to information. j Ct(v) represents the set of tags that user j assigns to information. i )∩Ct(v j Let ) represent the set of common labels of user i and user j, and || represent the value operation, which gives the number of common labels of user i and user j.
[0073] In this embodiment, user group information, such as the number of groups a user participates in, the number of public groups among users, the size of public groups among users, and the number of public tags among users, is used as user attribute features to add edge attributes and vertex attributes to the user relationship heterogeneous graph, thereby constructing a multi-attribute heterogeneous graph of user relationships. In this way, the multi-attribute heterogeneous graph of user relationships constructed based on user group information can effectively respond to the dynamic changes in user interests, thereby effectively and accurately discovering positive user relationships. Furthermore, based on the positive relationship users corresponding to the discovered target users, it can further assist in realizing timely and accurate information recommendation.
[0074] In other embodiments of the present invention, based on the above embodiments, the calculation of the relationship score between the target user and other users based on the relationship between vertices in the user relationship multi-attribute heterogeneous graph includes:
[0075] Based on the logistic regression model, the user relationship score calculation model is obtained by fusing features from the user's attribute characteristics, including the number of groups the user participates in, the number of common groups among users, the size of common groups among users, and the number of common tags among users.
[0076] The relationship score between the target user and other users is calculated based on the user relationship score calculation model and the user attribute characteristics of the target user.
[0077] Generally, positive relationship chains are not bidirectional. This means that if user A considers user B as a positive relationship, user B may not consider user A as a positive relationship. Therefore, it is necessary to predict positive relationships for each user individually. In this embodiment, for a single user, the attribute features selected for positive relationship prediction include the number of groups the user participates in, the number of common groups among users, the size of common groups among users, and the number of common tags among users. By fusing these attribute features using a logistic regression model, a user relationship score calculation model can be obtained. After obtaining the user relationship score calculation model, the attribute features of the target user and any other user are used as input parameters. By calculating the user relationship score using the model, the relationship score between the target user and any other user can be obtained, thus better clarifying the target user's true interests and identifying the target user's positive relationship groups.
[0078] It is understood that in this embodiment, the higher the score of the relationship between the target user and other users, the greater the likelihood that the target user has a positive relationship with those other users.
[0079] Specifically, in this embodiment, the user relationship score calculation model is as follows:
[0080]
[0081] Among them, score(v i ,v j ) represents the score of the relationship between user i and user j, |Gr(v i | represents the number of groups that user i participates in, |Gr(v j )| represents the number of groups that user j participates in, Cg(v i ,v j ) represents the number of common groups between user i and user j, Cgs(v i ,v j ) k Ct(v) represents the size of the common group between user i and user j. i ,vj ) represents the number of common labels between user i and user j, and n represents the number of common groups between user i and user j, i.e., n = Cg(v i ,v j ).
[0082] In other embodiments of the present invention, based on the above embodiments, the step of obtaining the positive relationship group of the target user according to a preset scoring threshold and the relationship score between the target user and other users includes:
[0083] The relationship score between the target user and other users is compared with a preset score threshold, and users whose relationship score with other users exceeds the preset score threshold are grouped into the same group to form the target user's positive relationship group.
[0084] Since a higher relationship score between the target user and other users indicates a greater likelihood of a positive relationship between the target user and those other users, this embodiment compares the relationship score between the target user and other users with a preset score threshold. Users whose relationship scores exceed the preset score threshold are grouped into the same group to form a positive relationship group for the target user. Specifically, the score threshold is set according to the number of users in the positive relationship group of the target user as needed for the experiment or in practice.
[0085] like Figure 2 The diagram shown is a structural schematic of a preferred embodiment of a user-positive relationship discovery system for information recommendation according to the present invention. The system may include:
[0086] According to a second aspect of the present invention, the present invention also provides a user-positive relationship discovery system for information recommendation, the system comprising:
[0087] Heterogeneous graph construction module 201 is used to construct a multi-attribute heterogeneous graph of user relationships, wherein the multi-attribute heterogeneous graph of user relationships uses users and information as vertices and the lines connecting the vertices as edges, and the edges are used to represent the relationships between the vertices, including the relationships between users and the relationships between users and information.
[0088] With the rapid development of online social networks, people are more inclined to trust recommendations from users with positive relationships. Discovering user relationships, especially positive relationships, can effectively improve the accuracy of information recommendations. To better discover positive user relationships and thus assist in timely and accurate information recommendations, it is first necessary to construct a multi-attribute heterogeneous graph of user relationships. Heterogeneous graphs contain various types of nodes and relationships. Compared to homogeneous graphs containing only one type of node / relationship, heterogeneous graphs are more complex and contain more information. Therefore, this embodiment of the invention uses a multi-attribute heterogeneous graph of user relationships to analyze the connections between users, which can effectively respond to the dynamic changes in user interests and user relationships, thereby effectively capturing the dynamic changes in user needs. Specifically, in this multi-attribute heterogeneous graph of user relationships, users and information are used as vertices, and the lines connecting the vertices are used as edges. The edges represent the relationships between vertices, including relationships between users and relationships between users and information.
[0089] User relationship calculation module 202 is used to calculate the relationship score between the target user and other users based on the relationship between vertices in the user relationship multi-attribute heterogeneous graph.
[0090] After constructing the multi-attribute heterogeneous graph of user relationships, the relationships between users are analyzed using this graph. Since the relationship score between a target user and other users can accurately and objectively reflect the positiveness of the relationship, this embodiment calculates the relationship score between the target user and other users based on the relationships between vertices in the multi-attribute heterogeneous graph, thereby quantifying the positiveness of the relationship between the target user and other users.
[0091] The positive relationship group generation module 203 is used to obtain the positive relationship group of the target user based on the preset scoring threshold and the relationship score between the target user and other users;
[0092] Since the relationship score between the target user and other users can accurately and objectively reflect the positiveness of the relationship between the target user and other users, after calculating the relationship score between the target user and other users, it is necessary to obtain the target user's positive relationship group based on the preset score threshold and the relationship score between the target user and other users. That is, to judge other users with positive relationships with the target user based on the relationship score between the target user and other users, the method of setting a score threshold is used to judge other users with positive relationships with the target user, and all users judged to have positive relationships with the target user are grouped together to obtain the target user's positive relationship group.
[0093] The active relationship user determination module 204 determines the N users most similar to the target user from the active relationship group of the target user based on the TOP-N algorithm, and uses the N users most similar to the target user as active relationship users for information recommendation to the target user, where N is a positive integer.
[0094] After obtaining the target user's active relationship group, to reduce the computational load of information recommendation and achieve faster and more accurate recommendations, it is necessary to determine the N users most similar to the target user from the target user's active relationship group based on the TOP-N algorithm. This involves sorting the users in the target user's active relationship group from highest to lowest relationship score and selecting the top N users as the target user's most similar users. These N most similar users are then used as active relationship users for information recommendations to the target user. This further filters the users in the target user's active relationship group using the TOP-N algorithm, narrowing down the range of other users with positive relationships to the target user for information recommendations. It should be noted that N is specifically set as needed.
[0095] In summary, this embodiment provides a user positive relationship discovery system for information recommendation. First, a heterogeneous graph construction module 201 constructs a multi-attribute heterogeneous graph of user relationships, where users and information are vertices, and the lines connecting the vertices are edges representing the relationships between the vertices, including relationships between users and relationships between users and information. Then, a user relationship calculation module 202 calculates the relationship score between the target user and other users based on the relationships between the vertices in the multi-attribute heterogeneous graph. Next, a positive relationship group generation module 203 obtains the target user's positive relationship group based on a preset score threshold and the relationship score between the target user and other users. Finally, a positive relationship user determination module 204 uses the TOP-N algorithm to determine the N users most similar to the target user from the target user's positive relationship group, and uses these N users as the positive relationship users for information recommendation to the target user. This embodiment analyzes the relationships between users by constructing a multi-attribute heterogeneous graph of user relationships. It can effectively respond to the dynamic changes in user interests and user relationships, thereby effectively capturing the dynamic changes in user needs. Based on the relationships between users, it calculates the positive relationship groups of target users, thereby realizing the discovery of positive user relationships and assisting in timely and accurate information recommendation.
[0096] In other embodiments of the present invention, based on the above embodiments, the heterogeneous graph construction module is specifically used for: constructing a multi-attribute heterogeneous graph of user relationships.
[0097] A heterogeneous graph of user relationships is constructed by using users and information as vertices and the lines connecting the vertices as edges.
[0098] Based on the user's attribute characteristics, edge attributes and vertex attributes are added to the heterogeneous graph of user relationships to obtain the multi-attribute heterogeneous graph of user relationships. The user's attribute characteristics include the number of groups the user participates in, the number of common groups among users, the size of common groups among users, and the number of common tags among users. The tags refer to the labels that users assign to information.
[0099] Specifically, in this embodiment, the user relationship multi-attribute heterogeneous graph G = (V, E, Gr, T) is constructed, where V represents the vertex set, E represents the edge set, Gr represents the group set, and T represents the set of labels that users have made to the information.
[0100] Since the existence of user associations depends on users' activities on the network, the determination of the existence of user associations can be linked to the discrimination of many user characteristics (i.e., user attribute features). However, no single user attribute is sufficient to predict the existence of user associations. This embodiment of the invention uses the following user attribute features to describe user activities:
[0101] (1) Number of groups participated in by users
[0102] The more groups a user participates in, the more active that user is in the network, and therefore the more likely that user is to connect with other users and have more relationships. Specifically, in this embodiment, the number of groups that user i participates in is represented as |Gr(v i )|, where Gr(v i Let ) represent the set of groups that user i participates in, and || represent the value operation, which yields the number of groups that user i participates in. Specifically, the group refers to the community to which the user belongs in the network, calculated by the community detection algorithm.
[0103] (2) Number of public groups among users
[0104] Since a higher number of common groups among users indicates greater similarity, the number of common groups is an important feature for predicting positive relationships. Specifically, in this embodiment, the number of common groups between user i and user j is represented as Cg(v i ,v j ),in,
[0105] Cg(v i,v j )=|Gr(v i )∩Gr(v j )|
[0106] Among them, Gr(v i Gr(v) represents the set of groups that user i participates in. j Gr(v) represents the set of groups that user j participates in, ∩ represents the intersection operation, and Gr(v) represents the set of groups that user j participates in. i )∩Gr(v j ) represents the set of groups that users i and j both participate in, and | represents the value operation, which gives the number of groups that users i and j both participate in.
[0107] (3) Size of public groups among users
[0108] If two users join a small public group, the likelihood of a positive relationship between them is high. This is because a large public group is likely a user group for a popular product or a large organization where users have loosely coupled relationships. Specifically, the group size characterizes the number of users in the group; the more users in a group, the larger the group. In this embodiment, the public group size between user i and user j is represented as Cgs(v i ,v j ),in,
[0109] Cgs(v i ,v j =Sizeof(Gr(v i )∩Gr(v j )).
[0110] (4) Number of public tags among users
[0111] Two users in the same group who share many common tags (common tags refer to the same or similar labels applied to the same information by both users; for example, if two users simultaneously give the same or similar reviews of the same product, then this same or similar review is a common tag shared by the two users) are more similar. Therefore, more common tags mean that the two users have more similar interests and are more likely to have a positive relationship. Specifically, the number of common tags between user i and user j is denoted as Ct(v i ,v j ),in,
[0112] Ct(v i ,v j )=|Ct(vi )∩Ct(v j )|
[0113] Wherein, Ct(v i Ct(v) represents the set of tags that user i has assigned to information. j Ct(v) represents the set of tags that user j assigns to information. i )∩Ct(v j Let ) represent the set of common labels of user i and user j, and | | represent the value operation, which gives the number of common labels of user i and user j.
[0114] In this embodiment, user group information, such as the number of groups a user participates in, the number of public groups among users, the size of public groups among users, and the number of public tags among users, is used as user attribute features to add edge attributes and vertex attributes to the user relationship heterogeneous graph, thereby constructing a multi-attribute heterogeneous graph of user relationships. In this way, the multi-attribute heterogeneous graph of user relationships constructed based on user group information can effectively respond to the dynamic changes in user interests, thereby effectively and accurately discovering positive user relationships. Furthermore, based on the positive relationship users corresponding to the discovered target users, it can further assist in realizing timely and accurate information recommendation.
[0115] In other embodiments of the present invention, based on the above embodiments, the user relationship calculation module calculates the relationship score between the target user and other users based on the relationship between vertices in the multi-attribute heterogeneous graph of user relationships. Specifically, for each user:
[0116] Based on the logistic regression model, the user relationship score calculation model is obtained by fusing features from the user's attribute characteristics, including the number of groups the user participates in, the number of common groups among users, the size of common groups among users, and the number of common tags among users.
[0117] The relationship score between the target user and other users is calculated based on the user relationship score calculation model and the user attribute characteristics of the target user.
[0118] Generally, positive relationship chains are not bidirectional. This means that if user A considers user B as a positive relationship, user B may not consider user A as a positive relationship. Therefore, it is necessary to predict positive relationships for each user individually. In this embodiment, for a single user, the attribute features selected for positive relationship prediction include the number of groups the user participates in, the number of common groups among users, the size of common groups among users, and the number of common tags among users. By fusing these attribute features using a logistic regression model, a user relationship score calculation model can be obtained. After obtaining the user relationship score calculation model, the attribute features of the target user and any other user are used as input parameters. By calculating the user relationship score using the model, the relationship score between the target user and any other user can be obtained, thus better clarifying the target user's true interests and identifying the target user's positive relationship groups.
[0119] It is understood that in this embodiment, the higher the score of the relationship between the target user and other users, the greater the likelihood that the target user has a positive relationship with those other users.
[0120] Specifically, in this embodiment, the user relationship score calculation model is as follows:
[0121]
[0122] Among them, score(v i ,v j ) represents the score of the relationship between user i and user j, |Gr(v i | represents the number of groups that user i participates in, |Gr(v j )| represents the number of groups that user j participates in, Cg(v i ,v j ) represents the number of common groups between user i and user j, Cgs(v i ,v j ) k Ct(v) represents the size of the common group between user i and user j. i ,v j ) represents the number of common labels between user i and user j, and n represents the number of common groups between user i and user j, i.e., n = Cg(v i ,v j ).
[0123] In other embodiments of the present invention, based on the above embodiments, the positive relationship group generation module is specifically used to: obtain the positive relationship group of the target user based on a preset score threshold and the relationship score between the target user and other users.
[0124] The relationship score between the target user and other users is compared with a preset score threshold, and users whose relationship score with other users exceeds the preset score threshold are grouped into the same group to form the target user's positive relationship group.
[0125] Since a higher relationship score between the target user and other users indicates a greater likelihood of a positive relationship between the target user and those other users, this embodiment compares the relationship score between the target user and other users with a preset score threshold. Users whose relationship scores exceed the preset score threshold are grouped into the same group to form a positive relationship group for the target user. Specifically, the score threshold is set according to the number of users in the positive relationship group of the target user as needed for the experiment or in practice.
[0126] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.
[0127] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.
[0128] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.
[0129] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for discovering positive user relationships for information recommendation, characterized in that, The method includes: Construct a multi-attribute heterogeneous graph of user relationships, wherein the multi-attribute heterogeneous graph of user relationships uses users and information as vertices and the lines connecting the vertices as edges, the edges are used to represent the relationships between the vertices, the relationships include the relationships between users and the relationships between users and information; Calculate the relationship score between the target user and other users based on the relationships between vertices in the aforementioned multi-attribute heterogeneous graph of user relationships. Based on preset scoring thresholds and the score of the target user's relationship with other users, a positive relationship group for the target user is obtained; Based on the TOP-N algorithm, N users most similar to the target user are determined from the target user's active relationship group, and the N users most similar to the target user are selected as active relationship users for information recommendation to the target user, where N is a positive integer; in, The construction of the multi-attribute heterogeneous graph of user relationships includes: A heterogeneous graph of user relationships is constructed by using users and information as vertices and the lines connecting the vertices as edges. Based on the user's attribute characteristics, edge attributes and vertex attributes are added to the user relationship heterogeneous graph to obtain the user relationship multi-attribute heterogeneous graph. The user's attribute characteristics include the number of groups the user participates in, the number of common groups among users, the size of common groups among users, and the number of common tags among users. The tags refer to the labels that users assign to information. The calculation of the relationship score between the target user and other users based on the relationship between vertices in the multi-attribute heterogeneous graph of user relationships includes: Based on the logistic regression model, the user relationship score calculation model is obtained by fusing features from the user's attribute characteristics, including the number of groups the user participates in, the number of common groups among users, the size of common groups among users, and the number of common tags among users. The relationship score between the target user and other users is calculated based on the user relationship score calculation model and the user attribute characteristics of the target user. The process of obtaining the target user's positive relationship groups based on a preset scoring threshold and the relationship score between the target user and other users includes: The relationship score between the target user and other users is compared with a preset score threshold, and users whose relationship score with other users exceeds the preset score threshold are grouped into the same group to form the target user's positive relationship group.
2. The method for discovering positive user relationships for information recommendation according to claim 1, characterized in that, The user relationship score calculation model is as follows: ； in, Indicates user With users Relationship score Indicates user Number of groups participated in Indicates user Number of groups participated in Indicates user With users The number of public groups between them Indicates user With users The size of the public groups between them Indicates user With users The number of common tags between them Indicates user With users The number of common groups between them, i.e. .
3. A user-positive relationship discovery system for information recommendation, characterized in that, The system includes: The heterogeneous graph construction module is used to construct a multi-attribute heterogeneous graph of user relationships, wherein the multi-attribute heterogeneous graph of user relationships uses users and information as vertices and the lines connecting the vertices as edges. The edges are used to represent the relationships between the vertices, including the relationships between users and the relationships between users and information. The user relationship calculation module is used to calculate the relationship score between the target user and other users based on the relationship between vertices in the multi-attribute heterogeneous graph of user relationships. The positive relationship group generation module is used to generate positive relationship groups for a target user based on a preset scoring threshold and the score of the relationship between the target user and other users. The active relationship user identification module determines the N users most similar to the target user from the active relationship group of the target user based on the TOP-N algorithm, and uses the N users most similar to the target user as active relationship users for information recommendation to the target user, where N is a positive integer; in, The heterogeneous graph construction module is specifically used to construct a multi-attribute heterogeneous graph of user relationships when performing the following operations: A heterogeneous graph of user relationships is constructed by using users and information as vertices and the lines connecting the vertices as edges. Based on the user's attribute characteristics, edge attributes and vertex attributes are added to the user relationship heterogeneous graph to obtain the user relationship multi-attribute heterogeneous graph. The user's attribute characteristics include the number of groups the user participates in, the number of common groups among users, the size of common groups among users, and the number of common tags among users. The tags refer to the labels that users assign to information. When the user relationship calculation module performs the calculation of the relationship score between the target user and other users based on the relationship between vertices in the multi-attribute heterogeneous graph of user relationships, the specific user is: Based on the logistic regression model, the user relationship score calculation model is obtained by fusing features from the user's attribute characteristics, including the number of groups the user participates in, the number of common groups among users, the size of common groups among users, and the number of common tags among users. The relationship score between the target user and other users is calculated based on the user relationship score calculation model and the user attribute characteristics of the target user. The positive relationship group generation module is specifically used to generate positive relationship groups for a target user based on a preset score threshold and the relationship score between the target user and other users: The relationship score between the target user and other users is compared with a preset score threshold, and users whose relationship score with other users exceeds the preset score threshold are grouped into the same group to form the target user's positive relationship group.
4. The user-positive relationship discovery system for information recommendation according to claim 3, characterized in that, The user relationship score calculation model is as follows: ； in, Indicates user With users Relationship score Indicates user Number of groups participated in Indicates user Number of groups participated in Indicates user With users The number of public groups between them Indicates user With users The size of the public groups between them Indicates user With users The number of common tags between them Indicates user With users The number of common groups between them, i.e. .