Cluster Head Selection Algorithm For Wireless Sensor Networks Using Machine Learning

Abstract

A Wireless Sensor Network (WSN) is a group of hardware sensors linked together over a wireless network. The sensors are segregated into clusters where every cluster has a cluster head whose job is to collect and transmit data back to the sink node (base station). A sensor node performs activities like data capturing, processing and transfer, which consumes energy. Since these nodes are usually deployed in areas not easily accessible by humans, battery life is one of the key aspects to work on in WSNs. Choosing the appropriate node as a cluster head improves the energy efficiency of the network. This paper proposes a K-Means Energy Efficient Cluster Head Selection Algorithm (KE-CHSA) for WSN where machine learning is applied to form clusters and elect cluster heads. We propose an equation which dynamically changes the number of clusters every round, based on the number of sensor nodes alive and a density parameter. Our algorithm elects the best-suited node as the cluster head considering the energy remaining and its distance from the other nodes. KE-CHSA outperforms the traditional LEACH [21] and C-LEACH [22] by improving the lifetime of WSNs.