A Core-based Community Detection Algorithm for Networks

Community detection is now playing a significant role in the discovery of underlying structures of social networks. This problem has been proved to be very hard and not been satisfactorily solved yet. Most of the algorithms proposed so far tend to maximize the number of intra-cluster edges, but ignore the importance of the core nodes within clusters. In contrast, this paper proposes a core-based algorithm that makes use of these core nodes. It first computes the core values of each vertex, and then gradually chooses the vertex with the maximum core value and performs a cluster expansion based on a structural similarity measurement. Evaluation using both real and synthetic datasets demonstrates that our method is not only efficient but also effective.