IncOrder: Incremental density-based community detection in dynamic networks

Abstract

In this paper, an incremental density-based clustering algorithm IncOrder is proposed for detecting communities in dynamic networks. It consists of two separate stages: an online stage and an offline stage. The online stage maintains the traversal sequence of a network and the offline stage extracts communities from the sequence. Based on a symmetric measure core-connectivity-similarity between pairs of adjacent nodes, the online stage builds an index structure, called core-connected chain, for dynamic networks. Since the slight change of a network has a very limited impact on its cluster chain, the chain of a dynamic network can be efficiently preserved. The offline stage extracts all possible density-based clustering results for all similarity thresholds from the chain. By maximizing a modularity function, the proposed method can automatically select the parameter of similarity threshold. Experimental results on a large number of real-world and synthetic networks show that the proposed method achieves high accuracy and efficiency.