Research On Community Discovery Based On K-shell

With the rapid development of science and information technology in human society,the research on objective things and scientific problems has not only focused on individual development and change,but also based on group,system and other comprehensive information bodies.In particular,community structure,as an important property of social network,can effectively reflect the behavior characteristics of individual members and their relationship with other individuals.The high quality community discovery can help to analyze and understand the internal law of social network,and it plays an important theoretical and research value on social network research.Firstly,it improves to an existing based on weighted aggregation coefficient method of set pair connection degree of community found VSFCM algorithm.Due to the clustering of the original algorithm is excessive dependence on the similarity,some nodes in the initial stage of polymerization merge early.Because there is no backtracking capability in hierarchical clustering algorithm,the result of final community discovery is unreasonable.In order to improve this problem,the edge clustering coefficient and similarity are used as the criteria of clustering discrimination.Secondly,in order to solve the problem of high time efficiency cost of VSFCM algorithm,the k-shell decomposition method was introduced to solve the problem.The network is roughly divided by the K-shell method,and then the initial clustering is carried out according to the K-shell value from large to small.At last,the similarity measure between the community nodes is analyzed with the similarity measure method of the weighted clustering coefficient to analyze the community nodes,and the KPCM algorithm and the KPCMV algorithm are proposed.The KPCM algorithm focuses on the superiority of the community structure.KPCMV algorithm focuses on the closer to the real network partition.Finally,coding implements the above three methods and applies them to real social networks: Zachary karate club network,dolphin network and American soccer network.The result of the experiment is in accordance with the actual situation.