Integrated structure investigation in complex networks by label propagation

TL;DR

This paper introduces LINSIA, a label propagation-based algorithm that comprehensively uncovers hierarchical, overlapping communities, hubs, and outliers in complex networks without prior knowledge, validated on synthetic and real data.

Contribution

LINSIA is a novel, parameter-free method that reveals multiple network structure features simultaneously, improving over existing partial solutions.

Findings

LINSIA effectively detects hierarchical and overlapping communities.

LINSIA identifies hubs and outliers accurately.

LINSIA outperforms state-of-the-art algorithms in experiments.

Abstract

The investigation of network structure has important significance to understand the functions of various complex networks. The communities with hierarchical and overlapping structures and the special nodes like hubs and outliers are all common structure features to the networks. Network structure investigation has attracted considerable research effort recently. However, existing studies have only partially explored the structure features. In this paper, a label propagation based integrated network structure investigation algorithm (LINSIA) is proposed. The main novelty here is that LINSIA can uncover hierarchical and overlapping communities, as well as hubs and outliers. Moreover, LINSIA can provide insight into the label propagation mechanism and propose a parameter-free solution that requires no prior knowledge. In addition, LINSIA can give out a soft-partitioning result and depict the degree of overlapping nodes belonging to each relevant community. The proposed algorithm is validated on various synthetic and real-world networks. Experimental results demonstrate that the algorithm outperforms several state-of-the-art methods.