Cider: Highly Efficient Processing of Densely Overlapped Communities in Big Graphs

Abstract

As one of the most fundamental operations in graph analytics, community detection is to find groups of vertices that are more densely connected internally than with the rest of the graph. However, the detection of densely overlapped communities in big graphs is extremely challenging due to high time complexity. In this paper, we propose an effective and efficient graph algorithm called Cider to detect densely overlapped communities in big graphs. The intuition behind our algorithm is to exploit inherent properties of densely overlapped communities, and expand the community by minimizing its conductance. To make Cider more efficient, we extend the algorithm to expand the community more quickly by merging vertices in batches. We explicitly derive the time complexity of our algorithm and conclude that it can be implemented in near-linear time. Besides, we also implement a parallelized version of Cider to further improve its performance. Experimental results on real datasets show that our algorithms outperform existing approaches in terms of Flake Out Degree Fraction (FODF) and $F_{1}\ Score$.