Overlapping communities detection of social network based on hybrid C-means clustering algorithm

As an important part of social computing, community detection has been attached to more and more importance in social network analysis. Overlapping communities detection, one of significant topics, is benefit to understand properties of knowledge sharing organization in social network. Because of uncertainties inherent in knowledge sharing organization, good results are hard to gain by using traditional community detection technologies. Through complement of both fuzzy sets and rough sets, this paper proposed a novel hybrid clustering method, which uses fuzzy partitioning technique to replace a traversal search method for discovering overlapping community structures. The final representation leads to an efficient description of overlapping regions among communities, as well as uncertainties in class boundaries. Meanwhile, with considering both local and global structural features of knowledge sharing organization in complex networks, a meaningful similarity measure for each pair of objects is designed. As a result, our proposed method can effectively and efficiently detect communities whose boundaries are not easily separated from each other. Further, experimental results on synthetic complex networks and real-world networks demonstrate that the proposed method works well on detecting overlapping community structures in a knowledge sharing organization of complex networks.