A method for identifying key users on large-scale social networks

Abstract

The present invention discloses a method for identifying key users on large-scale social networks, comprising the following steps: Step 1: Obtaining a k-hop subgraph of each user node based on the social network structure; Step 2: Obtaining the betweenness centrality of each user node and the upper and lower bounds of the betweenness centrality through hierarchical calculation on the k-hop subgraph of each user node; Step 3: Identifying key users through comparative judgment based on the betweenness centrality of each user node and the upper and lower bounds of the betweenness centrality, thereby obtaining a set of key users. The present invention can quickly locate key users who play an important role in information dissemination in large-scale social networks, providing support for social network management and public opinion prediction and control.

Description

Technical Field

[0001] The present invention relates to the field of social network user identification methods, and in particular to a method for identifying key users on a large-scale social network. Background Art

[0002] Identifying key users in large-scale social networks is a crucial task, crucial for the operation and development of social networking platforms. With the rapid development of the internet, social networks have become an integral part of people's daily lives, hosting a wide range of information dissemination, social interactions, and commercial activities. Key users are often important mediators and centers of influence within social networks. With extensive social network connections and influence, they can rapidly disseminate information and guide discussion and conversation. Therefore, identifying the behavioral patterns and influence of key users in such large and complex networks is crucial.

[0003] In network analysis, betweenness centrality is an important node metric. Metrics based on betweenness centrality effectively leverage the structural characteristics of users on a network to identify key users with significant influence in the dissemination process. However, large-scale social networks, such as Weibo, typically have billions of user nodes. Calculating betweenness centrality on such vast networks is prohibitively expensive. Furthermore, in some scenarios, such as when the network diameter is very large, betweenness centrality cannot effectively capture the influence and importance of user nodes.

[0004] Therefore, it is necessary to design a method for identifying key users on large-scale social networks based on betweenness centrality to solve the above problems. Summary of the Invention

[0005] The present invention provides a method for identifying key users on large-scale social networks. The method calculates key user nodes on large-scale social networks with relatively low computational overhead in various scenarios, and can quickly identify user nodes that play an important role in information dissemination in the network, so as to solve the problems of high computational cost and poor accuracy in identifying key users based on betweenness centrality in existing social networks, and provide support for subsequent prediction and control tasks of public opinion on social networks.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] A method for identifying key users on a large-scale social network includes the following steps:

[0008] Step 1: According to the social network structure G, obtain the K-hop subgraph corresponding to each user node i on the social network structure G

[0009] Step 2: K-hop subgraph of each user node i Through layered calculation, we get the K-jump subgraph The distance d(s,t) between user nodes s and t in the node pair (s,t) formed by any two user nodes s and t except user node i is obtained, and the K-hop subgraph is obtained according to the user node distance d(s,t) of the node pair (s,t). The local betweenness centrality b of the middle node pair (s, t) with respect to user node i i The value range of (s,t);

[0010] Then, according to the local betweenness centrality b i (s, t) calculation results, the node pair (s, t) is divided into three categories, among which the local betweenness centrality b i Node pairs whose (s, t) calculation results are 0 are classified as the first category, and the local betweenness centrality b i The node pairs whose (s, t) calculation result is 1 are classified as the second type, and the local betweenness centrality b i Node pairs whose (s, t) values ​​range from 0 to 1 are classified as the third category, and the number of node pairs in the first category H0, the number of node pairs in the second category H1, and the number of node pairs in the third category H2 are counted;

[0011] Then, according to the number of the first type of node pairs H0, the K-hop subgraph of each user node i is combined The total number of nodes n in the graph is obtained, and each user node i is in the K-hop subgraph The upper bound C of betweenness centrality in u ; And according to the number of the second type of node pairs H1, we can get the K-hop subgraph of each user node i The lower bound C of betweenness centrality in l ;

[0012] Step 3: Count the upper bounds of the betweenness centrality of all user nodes in the social network structure G in their respective K-hop subgraphs, and sort them in descending order according to the upper bounds. Calculate the betweenness centrality of the nodes in sequence and put them into the target node set S until the set size reaches the predetermined target value; then, take the smallest betweenness centrality b min (v) as a base value for judgment;

[0013] Continue to sequentially increase the betweenness centrality upper bound C of each user node i u Respectively with the judgment base value b min (v) Perform the initial comparison and judgment, if the upper bound of the betweenness centrality of the current user node i is C u Less than or equal to the judgment base value b min (v), then the current user node i is considered not to be a key user node; if the betweenness centrality upper bound C of the current user node iu Greater than the judgment base value b min (v), then calculate the exact betweenness centrality b of the current user node i i , and the exact betweenness centrality b of the current user node i i and the judgment base value b min (v) Conduct further comparison and judgment. The process of further comparison and judgment is as follows:

[0014] If the exact betweenness centrality of the current user node i is b i Less than or equal to the judgment base value b min (v), then the current user node i is considered not to be a key user node;

[0015] If the exact betweenness centrality of the current user node i is b i Greater than the judgment base value b min (v) If the current user node i is considered as a key user node, the current user node i is placed in the target node set S, and b is deleted from S. min (v) Corresponding node and updating b min (v).

[0016] In the further step 1, the social network structure G is stored in the form of an adjacency list, and a K-hop subgraph in the form of an adjacency list of each user node i is obtained according to the social network structure G in the form of an adjacency list.

[0017] In the further step 2, according to the K-hop subgraph of each user node i The distances from the remaining user nodes to the user node i except the user node i are calculated, and the K-hop subgraph of each user node i is converted into The nodes in are divided into multiple node sets, and then the induced subgraph of each node set is obtained, that is, the K-hop subgraph for each user node i Hierarchical subgraphs.

[0018] In the further step 2, the user node distance of the node pair (s, t) in each hierarchical subgraph is calculated, and the minimum user node distance of the node pair (s, t) in each hierarchical subgraph is used as the same-layer distance d of the node pair (s, t). c (s,t);

[0019] Then based on the same-layer distance d of the node pair (s, t) c (s,t), K-hop subgraph of user node i Perform iterative query on the neighbors of node pair (s, t) and obtain the neighbor relationships of node pair (s, t) in K-hop subgraph. The distance between user nodes in d(s,t).

[0020] Furthermore, the Dijkstra algorithm is used to calculate the user node distance of the node pair (s, t) in each hierarchical subgraph.

[0021] Furthermore, in step 2, each user node i in the K-hop subgraph The upper bound C of betweenness centrality in u The calculation formula is as follows:

[0022]

[0023] Furthermore, in step 2, each user node i in the K-hop subgraph The lower bound C of betweenness centrality in l The calculation formula is as follows:

[0024] C l =|H1|.

[0025] In the further step 3, if the current user node i is further compared and judged to be a key user node, the node with the smallest betweenness centrality b at this time is deleted from the target node set S. min (v) The corresponding node; then, when making subsequent node judgments, the minimum betweenness centrality is obtained again from the target node set S as the judgment base value of the subsequent nodes.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] 1. This paper proposes to use the betweenness centrality of user nodes on the K-hop subgraph in social networks to characterize the importance of user nodes in large-scale social networks, solving the limitation of existing betweenness centrality in accurately characterizing the influence of user nodes in some scenarios.

[0028] 2. This paper proposes a method that divides a K-hop subgraph into layers based on distance, calculates the distance between each pair of points, and ultimately obtains the distance between all pairs of points through same-layer distances and iterative cross-layer queries. This method avoids directly calculating the distances of all pairs of points on the complete graph structure, reducing computational overhead and increasing the method's scalability.

[0029] 3. This paper proposes a method to prune the target node search space by roughly calculating the upper bound of betweenness centrality from point pair classification. This method can significantly reduce the number of nodes that need to accurately calculate betweenness centrality, effectively reducing computational overhead. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 Flowchart for the proposed key user identification.

[0031] Figure 2 The overall framework diagram of the proposed key user identification. DETAILED DESCRIPTION

[0032] The present invention will be further described below with reference to the accompanying drawings and examples.

[0033] like Figure 1 、 Figure 2 As shown, this embodiment discloses a method for identifying key users on a large-scale social network, including the following steps:

[0034] Step 1: Assume that the user node set in the social network structure G is V, and each user node in the user node set V is i. Then the social network structure G = (V, E), where E is the set of associated edges between user nodes. According to the social network structure G, obtain the K-hop subgraph corresponding to each user node i on the social network structure G. Then the K-hop subgraph of the user node set V

[0035] In this embodiment, the social network structure G is stored in the form of an adjacency list. Calculate the K-hop subgraph corresponding to each user node i on the social network structure G When , we start from user node i and mark the distances of all nodes in the social network structure G to user node i through breadth-first search. Then, we iteratively traverse the neighboring nodes recorded in the adjacency table from user node i. If we traverse to a marked node, we record the node in the adjacency table. If we traverse to an unmarked node, we skip it. Finally, we get a K-hop subgraph of each user node i stored in the form of an adjacency table.

[0036] Step 2: K-hop subgraph of each user node i Through layered calculation, we get the K-jump subgraph The distance d(s,t) between user nodes s and t in the node pair (s,t) formed by any two user nodes s and t except user node i is obtained, and the K-hop subgraph is obtained according to the user node distance d(s,t) of the node pair (s,t). The local betweenness centrality b of the middle node pair (s, t) with respect to user node i i The range of (s, t) and the upper and lower bounds of betweenness centrality C u and C l The specific process is as follows:

[0037] (S1) According to the K-hop subgraph of each user node i The distances from the remaining user nodes to the user node i except the user node i are calculated, and the K-hop subgraph of each user node i is converted into The nodes in are divided into K node subsets

[0038] (S2) For each node subset obtained in step (S1), mark the node elements in the node subset, and then traverse the neighbors of the nodes in the node subset in sequence. When the traversal reaches a marked point, record the point in the adjacency list of the induced subgraph corresponding to the node subset, otherwise skip it. In this way, the induced subgraph corresponding to each node subset is obtained.

[0039] Finally, according to step (S1) and step (S2), K node subsets are obtained by distance stratification. The corresponding induced subgraphs, that is, a total of K induced subgraphs As a K-hop subgraph for each user node i Hierarchical subgraphs.

[0040] (S3) Let the node pair (s, t) be any two user nodes s and t in each hierarchical subgraph. Apply the Dijkstra algorithm to each node in each hierarchical subgraph to calculate the node spacing between the user nodes s and t in the node pair (s, t) of each hierarchical subgraph.

[0041] Then, the user node distances of the node pairs (s, t) calculated in each hierarchical subgraph are counted, and the minimum user node distance is obtained as the same-layer distance d of the node pair (s, t) c (s,t).

[0042] (S4) K-hop subgraph of user node i In the example, the same-layer distance d of the node pair (s, t) is c (s, t), and the distances d(s) and d(t) from user nodes s and t to user node i, respectively. Perform iterative query on the neighbors of node pair (s, t), thereby obtaining the neighbors of node pair (s, t) in the K-hop subgraph. The user node distance d(s,t) in the equation can also be used to determine whether there is a shortest path that does not pass through user node i.

[0043] (S5) According to the user node distance d(s,t) of the node pair (s,t), we get the K-hop subgraph The local betweenness centrality b of the middle node pair (s, t) with respect to user node i i The value range of (s,t), b i The formal definition of (s,t) is as follows:

[0044]

[0045] Where: s,t represents the number of shortest paths from node s to node t, σ s,t(i) represents the number of shortest paths from node s to node t that pass through node i; π(s,t) represents the set of nodes on the shortest path from node s to node t. When d(s,t) is less than d(s)+d(t), b i (s, t) takes the value of 0; when d(s, t) is equal to d(s) + d(t) and there is a shortest path that does not pass through user node i in the iterative query, b i (s, t) ranges from 0 to 1; otherwise b i (s,t) takes the value of 1.

[0046] In this embodiment, based on the above steps (S1)-(S5), the K-hop subgraph of each user node i is obtained. The local betweenness centrality b of the middle node pair (s, t) with respect to user node i i After the value range of (s, t), according to the local betweenness centrality b i (s, t) value range, the node pairs (s, t) are divided into three categories. Among them, the local betweenness centrality b i Node pairs whose (s, t) calculation results are 0 are classified as the first category, and the local betweenness centrality b i The node pairs whose (s, t) calculation result is 1 are classified as the second type, and the local betweenness centrality b i Node pairs with (s, t) values ​​between 0 and 1 are classified as the third category, and the number of node pairs in the first category H0, the number of node pairs in the second category H1, and the number of node pairs in the third category H2 are counted. The three variables H0, H1, and H2 fully record the classification of all point pairs, among which the number of node pairs in the first category H0 and the number of node pairs in the second category H1 are used for subsequent calculations.

[0047] Finally, the number of first-class node pairs H0 is combined with the K-hop subgraph of each user node i The total number of nodes n in the graph is obtained, and each user node i is in the K-hop subgraph The upper bound C of betweenness centrality in u , the calculation formula is as follows:

[0048]

[0049] And, according to the number of the second type of node pairs H1, we can get the K-hop subgraph of each user node i. The lower bound C of betweenness centrality in l , the calculation formula is as follows:

[0050] C l =|H1|.

[0051] Step 3: Based on the upper bound of the betweenness centrality of all user nodes in the social network structure G obtained in step 2 in their respective K-hop subgraphs, identify key users. The process is as follows:

[0052] (A1) Calculate the upper bound of the betweenness centrality of all user nodes in the social network structure G in their respective K-hop subgraphs, sort them in descending order according to the upper bound, and accurately calculate the betweenness centrality of the nodes in sequence and put them into the target node set S until the set size reaches the predetermined target value. The minimum betweenness centrality b is obtained. min (v) As a base value for judgment.

[0053] (A2) Continue to sequentially set the upper bound C of the betweenness centrality of each user node i u , respectively with the judgment base value b min (v) Perform the initial comparison and judgment, if the upper bound of the betweenness centrality of the current user node i is C u Less than or equal to the judgment base value b min (v), the current user node i is considered not to be a key user node, the current user node i is discarded, and step (A2) is restarted to perform the initial comparison judgment on the next user node;

[0054] If the upper bound C of the betweenness centrality of the current user node i u Greater than the judgment base value b min (v), then calculate the exact betweenness centrality b of the current user node i i , and the exact betweenness centrality b of the current user node i i and the judgment base value b min (v) Conduct further comparative judgment. The further judgment process is as follows:

[0055] If the exact betweenness centrality of the current user node i is b i Less than or equal to the judgment base value b min (v), the current user node i is considered not to be a key user node, the current user node i is discarded, and step (A2) is restarted to perform the initial comparison judgment on the next user node;

[0056] If the exact betweenness centrality of the current user node i is b i Greater than the judgment base value b min (v) If the current user node i is considered as a key user node, the current user node i is placed in the target node set S, and b is deleted from S. min (v) Corresponding node and updating b min (v).

[0057] (A3) If step (A2) further compares and determines that the current user node i is a key user node, then delete the node with the smallest betweenness centrality b from the target node set S.min (v) The corresponding node, and update b according to S min (v).

[0058] (A4) Repeat steps (A1) to (A3) until all key user nodes are identified from all user nodes, thereby obtaining a key user set.

[0059] The preferred embodiments of the present invention are described in detail above with reference to the accompanying drawings. The embodiments described in the present invention are merely descriptions of the preferred embodiments of the present invention and do not limit the concept and scope of the present invention. The various specific technical features described in the above specific embodiments can be combined in any suitable manner unless there is any contradiction. Such combinations should also be regarded as the contents disclosed in this disclosure as long as they do not violate the concept of the present invention. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations.

[0060] The present invention is not limited to the specific details of the above-mentioned embodiments. Within the scope of the technical concept of the present invention and without departing from the design concept of the present invention, various modifications and improvements made to the technical solution of the present invention by those skilled in the art should fall within the scope of protection of the present invention. The technical contents for which protection is sought in the present invention have been fully recorded in the claims.

Claims

1. A method for identifying key users on a large-scale social network, characterized in that: The following steps are involved: Step 1: According to the social network structure G, obtain the K-hop subgraph corresponding to each user node i on the social network structure G Step 2: K-hop subgraph of each user node i Through layered calculation, we get the K-jump subgraph The distance d(s,t) between user nodes s and t in the node pair (s,t) formed by any two user nodes s and t except user node i is obtained, and the K-hop subgraph is obtained according to the user node distance d(s,t) of the node pair (s,t). The local betweenness centrality b of the middle node pair (s, t) with respect to user node i i The value range of (s,t); Then, according to the local betweenness centrality b i (s, t) calculation results, the node pair (s, t) is divided into three categories, among which the local betweenness centrality b i Node pairs whose (s, t) calculation results are 0 are classified as the first category, and the local betweenness centrality b i The node pairs whose (s, t) calculation result is 1 are classified as the second type, and the local betweenness centrality b i Node pairs whose (s, t) values ​​range from 0 to 1 are classified as the third category, and the number of node pairs in the first category H0, the number of node pairs in the second category H1, and the number of node pairs in the third category H2 are counted; Then, according to the number of the first type of node pairs H0, the K-hop subgraph of each user node i is combined The total number of nodes n in the graph is obtained, and each user node i is in the K-hop subgraph The upper bound C of betweenness centrality in u ; And according to the number of the second type of node pairs H1, we can get the K-hop subgraph of each user node i The lower bound C of betweenness centrality in l ; Step 3: Count the upper bounds of the betweenness centrality of all user nodes in the social network structure G in their respective K-hop subgraphs, and sort them in descending order according to the upper bounds. Calculate the betweenness centrality of the nodes in sequence and put them into the target node set S until the set size reaches the predetermined target value; then, take the smallest betweenness centrality b min (v) as a base value for judgment; Continue to sequentially increase the betweenness centrality upper bound C of each user node i u Respectively with the judgment base value b min (v) Perform the initial comparison and judgment, if the upper bound of the betweenness centrality of the current user node i is C u Less than or equal to the judgment base value b min (v), then the current user node i is considered not to be a key user node; if the betweenness centrality upper bound C of the current user node i u Greater than the judgment base value b min (v), then calculate the exact betweenness centrality b of the current user node i i , and the exact betweenness centrality b of the current user node i i and the judgment base value b min (v) Conduct further comparison and judgment. The process of further comparison and judgment is as follows: If the exact betweenness centrality b of the current user node i i Less than or equal to the judgment base value b min (v), then the current user node i is considered not to be a key user node; If the exact betweenness centrality b of the current user node i i Greater than the judgment base value b min (v) If the current user node i is considered as a key user node, the current user node i is placed in the target node set S, and b is deleted from S. min (v) Corresponding node and updating b min (v).

2. The method for identifying key users on a large-scale social network according to claim 1, characterized in that: In step 1, the social network structure G is stored in the form of an adjacency list. Based on the social network structure G in the form of an adjacency list, a K-hop subgraph in the form of an adjacency list of each user node i is obtained.

3. The method for identifying key users on a large-scale social network according to claim 1, characterized in that: In step 2, according to the K-hop subgraph of each user node i The distances from the remaining user nodes to the user node i except the user node i are calculated, and the K-hop subgraph of each user node i is converted into The nodes in are divided into multiple node sets, and then the induced subgraph of each node set is obtained, that is, the K-hop subgraph for each user node i Hierarchical subgraphs.

4. The method for identifying key users on a large-scale social network according to claim 3, characterized in that: In step 2, the user node distance of the node pair (s, t) in each hierarchical subgraph is calculated, and the minimum user node distance of the node pair (s, t) in each hierarchical subgraph is used as the same-layer distance d of the node pair (s, t). c (s,t); Then based on the same-layer distance d of the node pair (s, t) c (s,t), K-hop subgraph of user node i Perform iterative query on the neighbors of node pair (s, t) and obtain the neighbor relationships of node pair (s, t) in K-hop subgraph. The distance between user nodes in d(s,t).

5. The method for identifying key users on a large-scale social network according to claim 4, characterized in that: The Dijkstra algorithm is used to calculate the user node distance between node pairs (s, t) in each hierarchical subgraph.

6. The method for identifying key users on a large-scale social network according to claim 1, characterized in that: In step 2, each user node i in the K-hop subgraph The upper bound C of betweenness centrality in u The calculation formula is as follows:

7. The method for identifying key users on a large-scale social network according to claim 1, characterized in that: In step 2, each user node i in the K-hop subgraph The lower bound C of betweenness centrality in l The calculation formula is as follows: C l =|H1|。 8. The method for identifying key users on a large-scale social network according to claim 1, characterized in that: In step 3, if the current user node i is further compared and judged to be a key user node, the node with the smallest betweenness centrality b at this time is deleted from the target node set S. min (v) The corresponding node; then, when making subsequent node judgments, the minimum betweenness centrality is obtained again from the target node set S as the judgment base value of the subsequent nodes.

Images