Unstructured Data Clustering using Hybrid K-Means and Fruit Fly Optimization (KMeans-FFO) Algorithm

Abstract

K-Means is the worldwide admirable clustering technique in which the datasets are partitioned into several clusters to solve various real world clustering problems. K-Means is sensitive to selection of initial cluster centroid and facing the problem of local optimal convergence. These limitations can be removed by using optimization techniques to enhance the clustering quality. In this research, a hybrid KMeans-Fruit Fly Optimization (KMeans-FFO) technique is implemented to obtain optimal quality of clustering. The quality of clustering is compared with K-Means, KMeans-PSO and KMeans-ALO clustering techniques based on several performance metrics on three unstructured datasets. The calculated outputs and statistical analysis indicate that the proposed hybrid KMeans-FFO has been obtained better quality of results as compared to the K-Means, KMeacns-PSO and KMeans-ALO on the basis of intracluster distance, Purity Index, F-Measure and Standard Deviation.