The research of community detection based on Micro-blog users

This study mainly focus on getting fine-grained and standard community partitions of Micro-blog user network. First, through analyzing the characteristics of micro-blog users, we gave the definition of similarity among users based on users' following behaviors and the definition of attractiveness of users based on the number of followers and friends. Relying on these two definitions, we defined the attractiveness to express the strength of any two users' relationship. Second, we chose three community detection algorithms: CNM, GRACLUS and MLRMCL to detect the community structure. Three results were obtained accordingly. Third, we made effort to get one comprehensive result by insertion, howerever, the effect is not as good as we expected. Therfore, we used a special method to handle the three results and obtained a satisfactory comprehensive result. At last, we analyzed the performance of the process and results on two data sets: LFR testing network and Sina Micro-blog user network. The analysis shows that both the definition of attractiveness between users and the method for obtaining the comprehensive results work well.