Title: Hybrid enhanced shuffled bat algorithm for data clustering
Authors: Reshu Chaudhary; Hema Banati
Addresses: Department of Computer Science, New Academic Block, University of Delhi, Delhi 110007, India ' Dyal Singh College, Lodi Road, Delhi 110003, India
Abstract: Enhanced shuffled bat algorithm (EShBAT) is a recently proposed variant of bat algorithm (BA) which has been successfully applied for numerical optimisation. To leverage the optimisation capabilities of EShBAT for clustering, HESB, a hybrid between EShBAT, K-medoids and K-means is proposed in this paper. EShBAT works by dividing the population of bats into groups called memeplexes, each of which evolve independently according to BA. HESB improves on that by employing K-medoids and K-means to generate a rich starting population for EShBAT. It also refines the memeplex best solutions at the end of every generation by employing K-means algorithm. Both these modifications combined together produce an efficient clustering algorithm. HESB is compared to BA, EShBAT, K-means and K-medoids, over ten real-life datasets. The results demonstrate the superiority of HESB.
Keywords: enhanced shuffled bat algorithm; EShBAT; K-means; K-medoids; data clustering.
International Journal of Advanced Intelligence Paradigms, 2020 Vol.17 No.3/4, pp.323 - 341
Received: 16 Jan 2017
Accepted: 09 Aug 2017
Published online: 11 Sep 2020 *