A Fast and High Quality Approach for Overlapping Community Detection through Minimizing Conductance

Abstract

Communities play an important role in the field of graph structure, especially in domains of networks analysis. A community (also referred to as a cluster) is a dense subgraph of the whole graph with more links between its members than between its members to the outside nodes. Communities overlap when nodes in graph belongs to multiple communities. Overlapping community detection is developing in the years, and seed set expansion is one successful technique in the area. In this paper, we propose a fast and high quality method in overlapping community detection using seeds expansion approach. In particular, we use personalized PageRank scheme that optimized the conductance of each community around all the seeds previously found, and after the phase we further minimize the conductance community score by searching for better position for each node in the graph with a novel and efficient algorithm. Experimental results in real graphs with ground truth communities show that our approach outperforms other state-of-art overlapping community detection methods, in particular, it is about two orders of magnitude faster than most of the existing methods with better or similar quality.