A Novel Snowball-Chain Approach for Detecting Community Structures in Social Graphs

Community detection in social networks has been a widely explored problem to gain information about the dense groups of user favoring some particular idea or topic, or sharing common interests. Inspired from the process of snowball sampling, this paper presents a novel community detection approach, termed as snowball-chain (SbChain), for identifying communities in social networks. SbChain follows a bottom-up approach to find the most prominent nodes based on the degree of overlapping neighbors and clustering coefficient, that may form some cliques. The novelty of SbChain lies in a single overlapping hyperparameter requirement, λ, to merge snowballs to form a community. λ also helps in deciding the coarseness of communities to be identified from the social graph. The proposed approach is evaluated over different real-world and synthetic benchmark datasets and compared with some state-of-the-art methods in terms of the number of identified communities and their modularity.