Authors: Nahlah M. Shatnawi
Addresses: Department of Computer Science, Yarmouk University, Irbid, Jordan
Abstract: One of the most popular clustering algorithms is K- means cluster due to its simplicity and efficiency. Although clustering using K-means algorithm is fast and produces good results, it still has a number of limitations including initial centroid selection and local optima. The purpose of this research is to develop a hybrid algorithm that address k-means clustering limitations and improve its performance by finding optimal cluster centre. In this paper, Lévy-flights or Lévy motion is one of non-Gaussian random processes used to solve the initial centroid problem. Bees algorithm is a population-based algorithm which has been proposed to overcome the local optima problem, used along with its local memory to enhance the efficiency of K-means. The proposed algorithm applied to different datasets and compared with K-means and basic Bees algorithm. The results show that the proposed algorithm gives better performance and avoid local optima problem.
Keywords: data clustering; k-means clustering; Levy flight; bees algorithm; initial centroid selection; local optima; optimal cluster centre; Levy motion.
International Journal of Business Intelligence and Data Mining, 2017 Vol.12 No.1, pp.14 - 24
Available online: 03 Mar 2017 *Full-text access for editors Access for subscribers Purchase this article Comment on this article