An Effective Comparison of Graph Clustering Algorithms via Random Graphs

Abstract

graph clustering algorithms have been proposed in recent past researches, each algorithm having its own advantages and drawbacks. All these algorithms rely on a very different approach so it's really hard to say that which one is the most efficient and optimal if we talk in the sense of performance. It is really hard to decide that which algorithm is beneficial in case of highly complex networks like PPI networks which consist of thousands of nodes. The paper proposes an effective data comparison of RNSC (Restricted Neighbourhood Search Clustering) and MCL (Markov Clustering) algorithms based on Erdos-Renyi and Power-Law Distribution graphs. The basic parameters used for comparison are Edge Density, Run Time, Number of Nodes, Cluster Size and Singleton Cluster. Our approach is an effective one because firstly we have used two types of graph generators, Erdos-Renyi and Scaled-Free for generation of input graphs which are very much closer to the real input graphs and secondly we have generated input graphs having more than 1000 nodes, so in our approach we have used both the algorithms for clustering highly complex input graphs just like PPI networks. For comparison and analysis purpose we have collected data sets and generated some graphs based on these parameters. The proposed approach depicts which algorithm is best to be used for clustering such complex graphs and also some fields for extension if possible in both them. All graphs used in this thesis are unweighted and undirected.