Analysis of Communities Evolution in Dynamic Social Networks

Abstract

In this paper we present a framework to study evolution of communities in dynamic networks. A dynamic network is represented by a sequence of static graphs named as network snapshots.We introduce a distance measure between static graphs to study similarity among network snapshots and to detect outlier events. To find a detailed structure within each network snapshot we used a modularity maximization algorithm based on a fast greedy search extended with a random walk approach. Community detection often results in a different number of communities in different network snapshots. To make communities evolution studies feasible we propose a greedy method to match clustering labels assigned to different networks. The suggested framework is applied for analysis of dynamic networks built from real-world mobile datasets.