An hybrid clustering algorithm for optimal clusters in Wireless sensor networks

Abstract

Clustering is a technique that alleviates network congestion and increase the energy efficiency of the Wireless sensor network. Hierarchical clustering and k-means clustering are well established clustering algorithms. The convergence of hierarchical clustering is only based on the inconsistency criterion while the k-means clustering requires the number of clusters (k) which are user defined. It is only by trial and error better values for inconsistency and k can be obtained. This may not be optimal and clustering without optimal number of clusters leads to energy inefficient network. Hence, a new hybrid self decisive clustering technique based on Hierarchical Agglomerative Clustering and k-means algorithm is proposed here. The main objective of this algorithm is to arrive at an optimal number of clusters for a given set of nodes distributed over the geographical area. Added to this, the algorithm identifies the Cluster Head also. The proposed algorithm is implemented in Matlab and compared with the existing techniques. Results demonstrate the supremacy of the proposal with that of Hierarchical and k-means clustering.