Research On Community Detection Method Of Social Network Based On Label Influence

Recently,the rapid development of social applications,such as micro-blog and Facebook,promotes the research on social network,and one of the key research content is community detection.Community detection is important to understand the structure characteristics and the behavior patterns of real network.This paper focuses on the research of community detection,and proposes a label propagation algorithm based on the label influence.Firstly,based on the similarity of RA,an initial labels algorithm based on the link strength is proposed.Through combining the attributes of vertices and the topological of network,this algorithm defines the link strength to be the index of similarity between nodes.It is used in the process of initial labels to reduce the update in the first propagation.Secondly,to reduce the randomness in the process of label propagation and improve the stability of community detection,a label propagation algorithm based on label influence,named LS-LPA,is proposed.In this algorithm,the link strength of vertices is used in the process of label propagation,and the rate of labels which belong to the neighbor of current node is considered.In addition,on the basis of considering the relationship between the first level neighbors and the second level neighbors,the label influence based on the link strength of vertices is defined.It is used to complete the process of update,and the community structure of network is found.Thirdly,in order to solve the problem,such as small communities and the result of community is unstable,while the LS-LPA algorithm deals with large social network,a new community detection algorithm,called CSBCD,is proposed.In this algorithm,firstly,the initial of label is completed by using the link strength of vertices.Then,the link strength of communities is defined.On this basis,the clustering of nodes is completed by using the idea of hierarchical clustering,and the community detection is effectively achieved.At last,utilizing the modularity as the index,experiments on traditional social networks is performed to validate the correctness and effectiveness of proposed LS-LPA algorithm and CSBCD algorithm.