Detecting Overlapping Communities with Heuristic Expansion Method Based on Rough Neighborhood

Abstract

Detecting overlapping communities is a hot topic of graph mining research in recent years.The local expansion methods based on structural fitness function are one kind of overlapped community's detection methods,which could simultaneously discover overlapped and hierarchical structures.The paper presents the local expansion method based on rough neighborhood and introduces a kind of stability measure reflecting the inherent structural features of communities.Aimed at community drift and redundant calculation problems of general local expansion methods,a novel heuristic strategy about community seeds is adopted to reduce the computational complexity and improve the detection capability in local expansion process.On the condition of the structural fitness maximization,the cliques with the nodes of great degree are expanded as local seed communities.Then the proposed local expansion method merges the adjacent duplicate communities by community's stability measure and finally generates the natural overlapping communities.In the end,the experimental results on some real networks show that the rough neighborhood-based method could be more effective in detecting overlapping structures,while the algorithm has good scalability on a large scale data.