Uncovering overlapping community structure in static and dynamic networks

Yang Gao, Xiangzhan Yu, Hongli Zhang

Abstract: Community detection is an important research area in complex networks, for which the existing methods are often inaccurate or inefficient (1) at dealing with large real networks, (2) at dealing with dynamic networks. In this paper, we propose DIS, a localized algorithm for uncovering overlapping community structure in real large-scale networks, and ADIS, an adaptive community update method for dynamic networks. Experiments in large-scale real-world networks demonstrate that DIS achieves competitive performance among the baselines, in particular, DIS is over 100x faster than the global algorithms with better quality, and it obtains much more accurate communities than the local algorithms without utilizing priori information. Experiments in dynamic networks demonstrate that ADIS achieves competitive community structure compared to other dynamic methods.