A vertex similarity probability model for finding network community structure

Abstract

Most methods for finding community structure are based on the prior knowledge of network structure type. These methods grouped the communities only when known network is unipartite or bipartite. This paper presents a vertex similarity probability (VSP) model which can find community structure without priori knowledge of network structure type. Vertex similarity, which assumes that, for any type of network structures, vertices in the same community have similar properties. In the VSP model, "Common neighbor index" is used to measure the vertex similarity probability, as it has been proved to be an effective index for vertex similarity. We apply the algorithm to real-world network data. The results show that the VSP model is uniform for both unipartite networks and bipartite networks, and it is able to find the community structure successfully without the use of the network structure type. © 2012 Springer-Verlag.