Dynamic Social Network Analysis Using Latent Space Model and an Integrated Clustering Algorithm

Social network can be generally defined as a group of individuals who are connected by a set of relationships. A key characteristic of social networks is their continual change. However, the bulk of the analysis methods developed and popularized in the field of computer were static in that all information about the time that social interactions take place is discarded. Although recently there is some work on dynamic social network analysis, these studies have some weakness. In this paper, we proposed a new unified framework that enable to analysis of dynamic social network and that make use of the dynamic information of the social interactions. In contrast to most existing models, which only modeling relationships that change over time while not identifying the cluster, or vice versa. Our approach based on latent space and a two phases clustering method. It can accurately identify the cluster (and the core actors) at each time-step, and also observe the moving trend of actor´s positions (the change history of the structure of the social network). The experimental result on a real-life dataset shows a very encouraging analysis performance, and demonstrates the ability of the proposed framework on analysis of dynamic social network.