Similarity-Based Community Detection in Social Network of Microblog

As a kind of new-arising social media, Microblog has accumulated hundreds of millions of users in four years and the amount is still increasing quickly, because of its brevity, instantaneity and openness. The social influence of Microblog becomes more and more widely nowadays. It is significant to research for the community detection in the network of Microblog’s users both in theory and application. On one hand, most Microblog’s users are real persons and thus finding communities’ structure will help in revealing the behavior pattern of human being; on the other hand, Microblog’s users can be classified different groups based on the results from community detection, which will facilitate the accomplishment of targeted advertising. Given the features of Microblog, i.e., a directed network and the arbitrariness in establishing the following relation, this paper proposes a kind of similarity measure for users based on their behavior that following others and being followed by others, and defines its modularity function, then designs the community detection approach based on the modularity maximization inspired by the idea of fast greedy algorithm. Furthermore, this method has been generalized to Microblog network with tag information of users. Three real networks are processed in this approach. The results show that the approach proposed in this paper is more efficient on detecting the community structure of network of Microblog’s users, compared with Newman’s modularity maximization method, Infomap method and Walktrap method.