Research On Detecting Algorithms For Community Structures In Complex Networks Based On Intimacy

Community structure refers to the internal aggregation of complex networks.It is of great significance to do some research on cluster of complex networks for analysis of its network characteristics,such as,structure of topology,information spreading and mining.So,in recent years,more and more scholars have devoted themselves to the research area on community structure detection of complex networks.And a large number of community detection algorithms have been proposed.However,differences of characteristics in different networks result in various detection methods and certain problems.For instance,conditions on overlapping bound of overlapping community detection are excessively strict.Moreover,problems of information loss exist in previous researches on detecting directed-weighted networks and bipartite networks.And further research is badly needed.Therefore,three different methods based on the similarity to detect community structures in different networks have been proposed in this research.Main work and innovations are as follows:(1)A method of screening the initial network and the latest absorption function for detecting the overlapping community structures in networks is proposed.Through the selection of key nodes,the core area centered on them is formed.Then the initial community will be found by absorbing the closely connected nodes.And the final community will be detected by further absorption and expansion.A method of screening the initial network and the latest absorption function have been used to detect the overlapping community structures in network.The experimental results show that the algorithm can effectively detect the community structure of the network,and the time complexity of the algorithm is reasonable.(2)A method of detecting communities by asymmetric intimacy in directed-weighted network is proposed.Based on the study of the relationship between the nodes of the directed and weighted network,a new intimacy parameter is proposed,which can not only reflect the direct relationship between the nodes,but also the indirect relationship,meanwhile,this parameter can be extended to depict the relationship of communities.Cluster analysis is carried out by using the relationship between nodes,and the final result is selected according to the modularity optimization.The superiority of the proposed algorithm and rationality of intimacy parameters designed are demonstrated by the contrast experiment.(3)An algorithm for detecting communities in bipartite networks is proposed.Based on the analysis of characteristics of bipartite networks,relationships between the nodes ofsame types and different types are presented.Several sub-communities are formed by analyzing the relationships and nodes clustered.Then sub-communities will be further merged to form the final community structure through the relations between the societies.Experimental results verify the accuracy and reliability of parameters on intimacy for this algorithm and the proposed algorithm has better performance on detecting communities in bipartite networks.In this paper,three different community detection algorithms for different networks have been proposed,which have progressive relationship.An overlapping community structure detection algorithm by the core-vertex and method of detecting communities by asymmetric intimacy in directed-weighted network are applied to one-mode networks.And another algorithm is used to detect communities in bipartite networks.According to different networks,different evaluation parameters are used to analyze the results,which present the effectiveness and superiority of the algorithms.