Block Modelling and Learning for Structure Analysis of Networks with Positive and Negative Links

摘要

Currently, many community mining methods for signed networks with positive and negative links have been proposed, however, these methods can only efficiently find the community of signed networks and unable to find other structure, such as bipartite, multipartite and so on. In this study, we present a mathematically principled community mining method for signed networks. Firstly, a probabilistic model is proposed to model the signed networks. Secondly, a variational Bayesian approach is deduced to learn the proximation distribution of model parameters. In our experiments, the proposed method is validated in the synthetic and real-word signed networks. The experimental results show the proposed method not only can efficiently find communities of signed networks but also can find the other structure.