A Hadoop Based Approach for Community Detection on Social Networks Using Leader Nodes

Abstract

Community detection is the most common and growing area of interest in social and real-time network applications. In recent years, several community detection methods have been developed. Particularly, community detection in Local expansion methods have been proved as effective and efficiently. However, there are some fundamental issues to uncover the overlapping communities. The maximum methods are sensitive to enable the seeds initialization and construct the parameters, while others are insufficient to establish the pervasive overlaps. In this paper, we proposed the new unsupervised Map Reduce based local expansion method for uncovering overlapping communities depends seed nodes. The goal of the proposed method is to locate the leader nodes (seed nodes) of communities with the basic graph measures such as degree, betweenness and closeness centralities and then derive the communities based on the leader nodes. We proposed Map-Reduce based Fuzzy C- Means Clustering Algorithm to derive the overlapping communities based on leader nodes. We tested our proposed method Leader based Community Detection (LBCD) on the real-world data sets of totals of 11 and the experimental results shows the more effective and optimistic in terms of network graph enabled overlapping community structures evaluation.