Modeling homophily fusion on cascading for overlapping community detection in social networks

Abstract

Community detection has been well studied in network analysis, but the more realistic case of overlapping community detection remains a challenge. Particularly, existing overlapping community detection approaches fail to give good quality communities in diverse networks. This limitation arises because the existing approaches rely on network properties that lack sufficient insight into the underlying network structure. In this work, we propose Sangam, an overlapping community detection algorithm. It combines network properties and diffusion paths by leveraging both homophily and information diffusion to identify community structures. Extended modularity, overlapping modularity and NF1 are used to evaluate the quality and accuracy of identified communities. Additionally, the time complexity and runtime performance of the Sangam algorithm have been comprehensively analyzed. Experiments on 71 LFR benchmark networks and 12 real-world networks show that the Sangam algorithm outperforms state-of-the-art algorithms.