High-Quality Disjoint and Overlapping Community Structure in Large-Scale Complex Networks

TL;DR

This paper introduces an enhanced hierarchical clustering algorithm for large-scale complex networks that detects both disjoint and overlapping communities using dynamic structural similarity, validated through benchmark experiments.

Contribution

It presents an improved agglomerative clustering method incorporating dynamic structural similarity and extends it to identify fuzzy and overlapping communities.

Findings

Effective detection of disjoint communities in large networks

Successful extension to fuzzy and overlapping community detection

Competitive performance on benchmark graphs

Abstract

In this paper, we propose an improved version of an agglomerative hierarchical clustering algorithm that performs disjoint community detection in large-scale complex networks. The improved algorithm is achieved after replacing the local structural similarity used in the original algorithm, with the recently proposed Dynamic Structural Similarity. Additionally, the improved algorithm is extended to detect fuzzy and crisp overlapping community structure. The extended algorithm leverages the disjoint community structure generated by itself and the dynamic structural similarity measures, to compute a proposed membership probability function that defines the fuzzy communities. Moreover, an experimental evaluation is performed on reference benchmark graphs in order to compare the proposed algorithms with the state-of-the-art.