Clustering of scale free networks using a k-medoid framework

Abstract

Clustering is a very important topic in the field of pattern recognition and artificial intelligence. Also it has become popular in newer application areas like communication networking, data mining, bio-informatics, web mining, mobile computing etc. This article describes a network clustering technique based on PAM or k-medoid algorithm with appropriate modification. This algorithm works faster than the classical k-medoid based algorithms designed for networks and provides better results. A better final cluster structure is obtained as the sum of within cluster spreads, i.e., the clustering metric has improved drastically. The result has compared with those obtained by a graph k-medoid and a geodesic distance based (considering only highest degree nodes) network clustering algorithms. We have shown that the degree of a node is a significant contributor for better clustering.