Community detection combining topology and attribute information

摘要

Community structures detection is critical in the analysis of features and functions of complex networks. Traditional methods are mostly concerned with the topology information of networks when conducting community detection, and can only describe the community structures from one aspect. For a more comprehensive analysis of the network, there is often attribute information available and it is a good complement to topology information. In this paper, we propose two parameter-free models based on nonnegative matrix factorization (NMF for short), Topology and Attribute NMF (TANMF for short) and Topology and Attribute Symmetrical NMF (TASNMF for short), combining topology information and attribute information for community structures detection. In addition, the multiplicative update rules are designed and the convergence is proved. Systematic experiments on both the synthetic and the real networks demonstrate the effectiveness and efficiency of our methods.