Overlapping Community Detecting Based on Complete Bipartite Graphs in Micro-Bipartite Network Bi-Egonet

Abstract

Community detection has become a hot topic in complex networks. It plays an important role in information recommendation and public opinion control. Bipartite network, as a special complex network, reflects the characteristics of a kind of network in our life truly and objectively. Therefore, detecting community structure in bipartite networks is of great significance and has practical value. In this paper, we first introduce two concepts: a) a micro-bipartite network model Bi-EgoNet which can be used to analyze bipartite network from a micro view to reduce the complexity of structure in bipartite networks, and b) a complete bipartite graph that is a special bipartite graph with the indivisible property. Then, we propose a novel overlapping community detection algorithm based on a complete bipartite graph in micro-bipartite network Bi-EgoNet (CBGBEN), which combines advantages of both a complete bipartite graph and Bi-EgoNet to get an optimal community structure. The CBGBEN is evaluated on accuracy and effectiveness in several synthetic and real-world bipartite networks. The CBGBEN is compared with other excellent existing algorithms, and our experimental results demonstrated that CBGBEN is better at detecting overlapping community structure in bipartite networks.