Research On Community Detection Based On User Behavior Analysis

In the current context of networking and information technology,social network has become an important research object in studying people's social rules,interests and behavior characteristics.In recent years,more and more conferences and workgroups have taken community detection as one of the research questions and have achieved considerable results.However,in view of the complexity of social network structure and the diversity of content,traditional community detection algorithm is difficult to integrate various information and accurately divide social networks.Aiming at the characteristics of social networking,this paper studies community detection based on user behavior analysis and proposes a unified link and content(ULC)for heterogeneous social networks,which integrates all kinds of information.ULC algorithm can effectively analyze user behavior information in social network,transform it into user similarity measurement expression,and combine the network topology structure to get the final communities.Compared with traditional community detection algorithms that only use graph topology or only use node content information,ULC algorithm can make full use of information in social network,and effectively improve the accuracy.And the sampling step in the ULC algorithm reduces the time complexity of the algorithm.Experiments on the social network data set of the Douban show that,compared with several other existing algorithms,the F-score value index has been promoted.Compared with other community detection algorithms that integrate content and link,the consumption time is reduced.This paper studies the community detection in social networks.An algorithm that considers the topology structure of social networks and the user's behavioral preferences in the network for community detection is proposed.The main work is as follows:1.Most traditional community detection algorithms based on graph structure only focus on the direct link between nodes but based on graph connectivity,there is a certain connection between points that are not directly connected.In this paper,when dealing with the topology of graph,we use the neighborhood random walk distance to measure the closeness between nodes.This method can effectively measure the indirect relationship between nodes,and make full use of the information contained in network topology.The results are more scientific and reasonable.2.According to the characteristics of heterogeneous information network,this paper presents a method to evaluate the user relationship strength:topology analysis based on the relationship between user;user behavior on social networks brings connection between user and other components,using the combination of LDA model and tags to analyze information of network and measure the similarity of users by KL distance.The two are combined to assess user relationships and provide a basis for subsequent community detection.3.According to user similarity of links and user behavior information,a new network is formed after edge sampling,which has the same user nodes as the original social network,but with fewer edges.Traditional community detection algorithms can be applied to the framework,which makes the algorithm very flexible and the time complexity is low.This article will use an algorithm to fully consider the interaction between graph nodes for community detection.