Community detection in dynamic signed network: an intimacy evolutionary clustering algorithm

The tremendous development of community detection in dynamic networks have been witnessed in recent years. In this paper, intimacy evolutionary clustering algorithm is proposed to detect community structure in dynamic networks. Firstly, the time weighted similarity matrix is utilized and calculated to grasp time variation during the community evolution. Secondly, the differential equations are adopted to learn the intimacy evolutionary behaviors. During the interactions, intimacy between two nodes would be updated based on the iteration model. Nodes with higher intimacy would gather into the same cluster and nodes with lower intimacy would get away, then the community structure would be formed in dynamic networks. The extensive experiments are conducted on both real-world and synthetic signed networks to show the efficiency of detection performance. Moreover, the presented method achieves better detection performance compared with several better algorithms in terms of detection accuracy.