A systematic analysis of community detection in complex networks

Abstract

Numerous techniques have been proposed by researchers to uncover the hidden patterns of real-world complex networks. Finding a hidden community is one of the crucial tasks for community detection in complex networks. Despite the presence of multiple methods for community detection, identification of the best performing method over different complex networks is still an open research question. In this article, we analyzed eight state-of-the-art community detection algorithms on nine complex networks of varying sizes covering various domains including animal, biomedical, terrorist, social, and human contacts. The objective of this article is to identify the best performing algorithm for community detection in real-world complex networks of various sizes and from different domains. The obtained results over 100 iterations demonstrated that the multi-scale method has outperformed the other techniques in terms of accuracy. Multi-scale method achieved 0.458 average value of modularity metric whereas multiple screening resolution, unfolding fast, greedy, multi-resolution, local fitness optimization, sparse Geosocial community detection algorithm, and spectral clustering, respectively obtained the modularity values 0.455, 0.441, 0.436, 0.421, 0.368, 0.341, and 0.340..