Community Change Detection in Dynamic Networks in Noisy Environment

Sadamori Koujaku, Mineichi Kudo, Ichigaku Takigawa, Hideyuki Imai

Abstract: Detection of anomalous changes in social networks has been studied in various applications such as change detection of social interests and virus infections. Among several kinds of network changes, we concentrate on the structural changes of relatively small stationary communities. Such a change is important because it implies that some crucial changes have happened in a special group, such as dismiss of a board of directors. One difficulty is that we have to do this in a noisy environment. This paper, therefore, proposes an algorithm that finds stationary communities in a noisy environment. Experiments on two real networks showed the advantages of our proposed algorithm.