Research Of Community Discovery Algorithm Based On Complex Network And Display Analysis System

With the wide application of various social networks such as Weibo,WeChat and forum,people have generated a large amount of data in social networks.The data constitutes a complex virtual social network,which is the physical reflection of social reality.Using the computing technology to analyze and research the law and development trend of social relations and social behavior in social networks,it is of great significance to research and solve the real social problems related in sociology.Community detection is the basis of complex social network research.However,in recent years,researchers have found that overlapping community detection can better reflect the realistic significance of real social conditions.Label propagation algorithm is widely used and one of the typical algorithms in community discovery algorithm,however,the label propagation algorithm existed two problems of label redundancy phenomenon in the initialization phase,and output instability and inaccuracy caused by randomly updating nodes' order in the process of the label propagation.This paper proposes an overlapping community discovery algorithm to apply homogeneous network based on K-group and the influence of network node.In the initialization stage,the algorithm finds K-groups in the network and assigns the same labels to the nodes belonging to the same K-group to improve the problem of label redundancy.In the propagation phase,the node label is updated according to the sort policy by calculating the semi-local centrality of nodes.Experiments on several standard data sets and real Weibo data sets show that the proposed algorithm not only improves the stability of the algorithm,but also improves the quality of community division.Previous research mainly concentrated on community detection in homogeneous complex networks.However the social network shows heterogeneous characteristic because of the pluralistic society in the reality.In fact some social network have objects owning different types,and coexist a variety of relations between network nodes.Therefore,more complex social network is no longer a homogeneous network and should be a heterogeneous network.It is the current hot spot how to develop community detection in heterogeneous network.This paper argues that it is the important problem to depict intimacy relationship between objects in heterogeneous network.So in this paper,the random walk method is applied to calculate the closeness degree between different nodes in heterogeneous network,and then heterogeneous network is to remodel a homogeneous weighted network according to closeness degree.Because random walk algorithm fully considers the various relationships between different nodes of heterogeneous network,the remodeling weighted network not only effectively transform the structural relationship of heterogeneous networks,but also seek a solution of community detection for heterogeneous network.In this paper,firstly a heterogeneous network data sets are analyzed and constructed through DBLP literature data sets,and experiments is carried out to test the method of calculating node affinity of heterogeneous network and effect of community detection based on remodeling weighted network.In order to more clearly and effectively display the research contents about complex network,this paper designs and implements a display and analytic system of community detection.The display part of this system mainly shows a visualization display of the original network structure about complex social network in different data sets,and a visualization result of classic and proposed community detection algorithm on different data sets.The analytic part of this system is to show the application of social network,which realized the sentiment analysis of the Microblog hot issues and emotional tendency analysis of Weibo users on typical events.