Community detection based on unified Bayesian nonnegative matrix factorization

Community detection is of great importance to social networks analysis. It is widely acknowledged that the number of inner partitions (or dimensions) is related to the link information which exists in communities. But in fact, it is difficult to uncover the underlying community structure in traditional community detection approaches especially in real-world condition where rich attribute information exists. Therefore, how to integrate with link information and attribute information in community detection is a challenging problem. In this paper, we present a novel method based on Unified Bayesian Nonnegative Matrix Factorization: UBNMF, we develop a Bayesian approach for link and attribute information, and then we apply it into the Nonnegative Matrix Factorization (NMF) for community detection. Experiments on real data shows that the community detection approach has an ideal calculate result, the framework can easily detect the internal community structure, as well as can find the overlapping communities effectively, especially in complex networks.