Research Of Detecting Community Based On Label Propagation In Social Networks

As the research and development of complex networks, technology research and application of detecting community gradually became the key research direction of network analysis. Studying community detection is important to realize the social network structure and characteristics, and community detection methods include non-overlapping and overlapping community detection. However, the some vertexes may be belonged to multiple communities, overlapping community detection is more in line with the realistic requirements.Currently, the community detection algorithm based on label propagation has been widely used and in-depth researched. However, COPRA algorithm can effectively detect the overlap community structure from the networks, but the algorithm has some drawbacks which are strong randomness, poor robust, all the vertices are easily assigned to one community, and in this paper a new algorithm in which the network vertex label is updated according to the label entropic order, named the community overlap propagation algorithm based on label entropic order (COPRAE), is proposed to improve the accuracy and stability of detecting community. Meanwhile, in order to further improve the stability of COPRAE algorithm, if multiple labels are selected to a vertex, the algorithm uses the strategy of examining the labels of vertex itself. At last, the accuracy and stability of COPRAE algorithm is tested by using the synthetic networks and real networks.In everyday life, people constituted the different groups in social network because of different relationship during the people, and detecting groups from the networks is important implications for the practical affairs of life. Aimed at the BBS user network, We propose a weight algorithm based on interest similarity, our algorithm consider the relationship between users and the users’ interests to build user network, and then detect community from the user network. Our algorithm is able to effectively detect interest communities from the BBS user network.