Game theoretic approach on analysis of wireless sensor networks

Abstract

Wireless sensor networks have become increasingly popular due to their wide range of applications. K-means is a typical clustering algorithm for clustering of these wireless sensor networks, and it is widely used for grouping of large sets of data, owing to its ease of computation and implementation. However, due to the limitations and inaccuracy of the algorithm, two novel approaches for clustering of wireless sensor networks have been comprehensively analyzed in this thesis. In this report, improved methods for clustering of wireless sensor networks for a data set representing the information gathered by the sensor nodes have been provided. Also, the performance and comparison of these methods has been done, based on the experimental results. PCA based K-Means Algorithm proved to be a more efficient approach for clustering, as compared to regular K-Means Clustering. Furthermore, a comprehensive analysis of the regular KMeans Algorithm was carried out, as compared to the Game Theoretic Weighted K-Means Algorithm. Shapley Values were used to perform relative