A New Iterative Modularity-Based Method for Graph Clustering on Scalable Networks

Abstract

Due to the advancement of technology, modern networks such as social networks, citation networks, Web networks have been extremely large, reaching millions of nodes in a network. But most of the existing graph clustering algorithms can only tackle with small or medium-size networks. In this paper, we introduce a new method which can achieve high graph clustering quality for large scale networks by optimizing the modularity function. It is based on the iterative idea and takes good advantage of the exsiting multilevel local search heuristics. After introducing this modularity-based method, we evaluate its performance by applying it to several well-known network datasets. With a cost of more but acceptable time, it outperforms the best algorithms in the literature in the case of modularity optimization quality.