Label Propagation Based On Community Core For Community Detection Algorithm

In real life, we often solve some actual problems by modeling.Tthe network structured can be described by abstract objects and relationships,then find common features in objects. Community structure is an important feature in complex networks,which has dense nodes in network. The nodes in the community are dense,while between those communities are sparse.Finding community structure is helpful to analyse network structure and get the hidden characteristics in the network. In this paper,the work has been done about community detection as follows:(1)This paper introduces the background and the significance of community deteciton, then summarize a lot of algorithms use in community division.Among many community detection algorithms,label propagation algorithm is widely used for it is simple and rapid.But it has the problem of poor stability and weak robustness and so on.Therefore, we improve the initialization process of label propagation algorithm to improve the community quality.In this paper we proposed a label propagation algorithm based on community core for community detection,by calculating any two nodes’s k order common neighbor,we will make the most sililar nodes and it’s k order neighbor nodes as the initial community core.According the above process,we have some tight structure as the initial label of label propagation, and to assign these structures the initial community label.In a real network,experimental results show that the algorithm can improve the stability of the results.(2)The label propagation algorithm use in bipartite graph can’t get the ideal results. In order to get the ideal results,we propose a LPA algorithm based on local information use in bipartite graph to get the community detection. Bipartite graph is composed of two types nodes,and nodes in the same class has no connection. Considering it’s network structure is special,we get the weighted network by projection the bipartite graph.In this algorithm,we need to find the initial core communities in the weighted network.Therefore,the number and size of the initial core communities has imporant influence on the whole algorithm. The initial core communities are influencd by the threshold edge weight. Getting the appropriate weight threshold is contributed to get the proper initial core communities.Then finding community in bipartite graph by t label propagation algorithm.In the label processing,if a node has multiple optional labels,then it can choose all.In this way,this algorithm can not get the community structure,and also find the overlapping nodes.The experimental results show that,it not only improve the quality of community deteciton in bipartite graph.and also can get the overlapping community structure.