Noise-tolerance community detection and evolution in dynamic social networks

Dynamic complex social network is always mixed with noisy data and abnormal events always influence the network. It is important to track dynamic community evolution and discover the abnormal events for understanding real world. In this paper, we propose a novel algorithm Noise-Tolerance Community Detection (NTCD) to discover dynamic community structure that is based on historical information and current information. An updated algorithm is introduced to help find the community structure snapshot at each time step. One evaluation method based on structure and connection degree is proposed to measure the community similarity. Based on this evaluation, the latent community evolution can be tracked and abnormal events can be gotten. Experiments on different real datasets show that NTCD not only eliminates the influence of noisy data but also discovers the real community structure and abnormal events.