Hierarchical Community Discovery for Social Networks Based on Probabilistic Topic Model

The traditional community discovery algorithms are generally based on the link structure of a given social network, they lack of consideration of user's interests and the hierarchical structure of community. In this paper, a novel PAM (Pachinko Allocation Model) probabilistic generative model is proposed to detect latent hierarchical communities based on the user interests and their social relationships. The joint model of topic modeling and community discovery can capture the correlation among multiple communities and their hierarchical structure. Experiments on real-world dataset have confirmed the feasibility and effectiveness of the proposed algorithm.