Comparative Study of PSO-Based Hybrid Clustering Algorithms for Wireless Sensor Networks

Clustering is a task which creates groups depending upon the presence of similarity between the data objects. Many clustering algorithms exist, which are capable of creating well-defined clusters. One of the popular algorithms is K-means, which is generally used for data clustering where performance is dependable on initial state of centroid but have limitation of trapping in local optima. Besides K-means, K-harmonic means, and Fuzzy C-means are also popular algorithms used for data clustering but again they have the same limitation of trapping in local optima. So this creates problem while handling anomaly existing dataset in wireless sensor network. In this paper, an analysis of best suitable hybrid clustering algorithm is brought for a congregation of normal and anomalous dataset by using a stochastic tool Particle Swarm Optimization (PSO) by utilizing different sensor datasets. The results are encouraging in terms of best suitable fitness function and low computational time.