K-means Clustering and PSO Algorithm for Wireless Sensor Networks Optimization

Abstract

K means algorithm is one of the powerful and popular unsupervised machine learning algorithms, typically used in wireless sensor networks WSNs to separate the network into subnetwork or smaller networks known as clusters K.

The purpose of clustering is to reduce the amount of energy consumed in the network which results into improve the lifetime of the network, Determining the optimal number of K is the most challenging in WSN research area. In the first part of this paper the two most powerful methods El-bow, and Silhouette approaches are used to form clustering and implemented on three different types of real-world applications depending on the size of network. Extensive simulations show that Elbow method is more appropriate in small and medium sized networks compared to silhouette method which shows its robustness in large network due to a minimum amount of time is used to form subgroups. In the second part of this study low energy adaptive clustering hierarchy (LEACH) and Particle Swarm Optimization PSO based LEACH protocol is utilized on small sized network to validate the efficacy of the clustering used in the first part, The results show that PSO based LEACH protocol outperforms LEACH protocol in terms of energy consumption and number of packets sent when the nodes are communicating. PSO based LEACH protocol sends more packets and has fewer dead nodes, resulting in lower energy consumption.