A Community Detection Algorithm Considering Edge Betweenness and Vertex Similarity

Abstract

Community detection is an important topic in social network analysis. It is beneficial to understand the underlying structure of the network and extract useful information from it. Most existing community detection algorithms require a prior information or neglect peripheral vertices. In this paper, we propose a divisive community detection algorithm named CDBS (Community Detection considering edge Betweenness and vertex Similarity). First, the betweenness and similarity of the connected pair of vertices are calculated for all edges in the network. Secondly, the edges with relatively high betweenness and low similarity between connected pairs of vertices are identified by two thresholds ($\delta$ and $\theta$). Then these edges are removed from the network and the betweenness of the remaining edges are recalculated. This procedure is iterated until there is no more edge of which the betweenness is higher than $\delta$ and similarity is less than $\theta$. Finally, the proposed algorithm is validated in both synthetic and real-world networks. Experimental results demonstrate that CDBS is effective at detecting dense community structure with high accuracy and modularity, and it is time-efficient because of low computational complexity. Besides, CDBS can cope with the isolated cluster problem.