Partition Methods For Social Networkcommunity Based On Label Propagation

This research for complex network community detection is of important theoretical and realistic significance. It has not only attracted a large number of researchers from different subjects, and also been applied to identify terrorist organizations and mine Web community, and so on. Most community detection algorithms have high complexity, which cannot be used in large scale complex network.Aiming at this issue, in this paper, the asynchronous label propagation algorithm with nearly linear complexity is improved and developed. The traditional asynchronous label propagation algorithm has poor stability. Even a small scale network will be divided into many kinds of community structures, while some structures don’t conform to the definition of community. In order to solve the above problems, this paper firstly analyzes the possible reasons of the problems, and then put forward an improved asynchronous label propagation algorithm according to the node influence to determine the order of node updating. Furthermore, we respectively use the node closeness from overall consideration and the node importance which combines comprehensive global characteristics and local characteristics to measure the influence of the nodes in the network. The experimental results show that the algorithm based on node influence effectively improved the stability of the asynchronous label propagation algorithm.In addition, since online network data are effective data sources for method verification. Therefore, in this paper, we first introduce the webpage structure of yin shui si yuan BBS and Tianya community, and then realize the data acquisition of these two websites which will be applied to verify the algorithms we propose. The data we acquire have also been applied in the online network data analysis system developed by the national 973 project assumed by laboratory recently.