Community detection in dynamic network using dirichlet process

Community detection is a widely used method to extract useful information from social networks. Since many types of data are time-dependent, dynamic network clustering has drawn great attention in recent years. A good dynamic clustering approach should result in a smooth cluster evolution and determine the number of communities automatically. In this paper, we propose a Dirichlet Process based Dynamic Network Clustering Method using Temporal Dirichlet Process with stochastic block model, which is able to detect communities and meet requirements mentioned above. We did experiments on synthetic data and result shows our method is outperformed several state-of-the-art methods in both the accuracy of determining the number of clusters and the capability of resisting noisy data.