An Overlapping Community Detection Approach in Ego-Splitting Networks Using Symmetric Nonnegative Matrix Factorization

Overlapping clustering is a fundamental and widely studied subject that identifies all densely connected groups of vertices separates them from other in complex networks. However, most conventional algorithms extract modules directly the whole large-scale graph using various heuristics, resulting either high time consumption or low accuracy. To address this issue, we develop an overlapping community detection approach Ego-Splitting networks symmetric Nonnegative Matrix Factorization (ESNMF). It primarily divides network into many sub-graphs under premise preserving property, then extracts well-connected sub-sub-graph round each seed as prior information to supplement adjacent matrix, finally precise communities via nonnegative matrix factorization sub-network. Experiments on both synthetic real-world publicly available datasets demonstrate proposed outperforms state-of-the-art methods for