IMPROVED PERFORMANCE CLUSTERING USING MODIFIED K-MEANS ALGORITHM IN MOBILE ADHOC NETWORKS

Mobile Adhoc Network (MANET) is a self-configuring and dynamic network composed of mobile nodes that form a temporary network without any base station. The routes between source and destination change frequently which results in a lot of routing overhead, congestion in the network, and higher energy consumption of nodes. This paper proposes a way for cluster formation and selection of stable cluster heads. This is done by forming clusters of nodes using a modified K-means algorithm in which initial centroids are based on the geographic area of the network. Thereafter, the selection of cluster heads is done using the weighted multi-criterion acceptability formula. These ideas result in improved quality of service concerning load balancing, PDF, and Throughput as compared to clustering using K-means with random centroids.