A Novel Regression Based Clustering Technique for Wireless Sensor Networks

The objective of this paper is to develop a new adaptive iterative linear regression based clustering algorithm for wireless sensor network. According to this, the initial cluster is classified horizontally and vertically in parallel, each resulting in two sub-clusters. Of these two, the best is selected based on the proposed similarity index and with this selected cluster as reference, iteration continues until the convergence criterion ‘Delta’ is met. The similarity index is designed based on the intra cluster similarity and inter cluster dissimilarity. Delta is the difference between the similarity index of the current iteration and the previous iteration. The proposal is implemented in MATLAB and simulations are carried out under different network scenarios. The cluster quality is evaluated through external and internal indices using the Cluster Validity Analysis Platform tool. The cluster obtained by the proposal is studied and its quality is compared with the well-established k-means and hierarchical clustering. The performance indices confirm the supremacy of the proposal.