Combining Link and Content for Community Detection in Social Networks

摘要

Networks depicting social interactions between people have been studied for decades. Most of social networks are composed of multiple interactions and different types of components, while current analysis often extracts only part of the information. In this paper, we approach unified link and content(ULC) model which combining different links and contents analysis for community detection. Links from users to other components are analyzed with topic model and tags to get the similarity of users. We combine similarity graph and user relation topology diagram to get a new weighted graph. Community detection is done with this new graphic model. Experimental results on douban dataset demonstrate superiority of the proposed model over the state of the art.