Experimental Study of Variation Local Search Mechanism for Bee Algorithm Feature Selection

Abstract

The Bees Algorithm (BA) has been applied for finding the best possible subset features of a dataset. However, the main issue of the BA for feature selection is that it requires long computational time. This is due to the nature of BA combination search approach that exploits neighborhoods with random explorative. This situation creates unwanted suboptimum solution(s) leading to the lack of accuracy and longer processing time. A set of different local neighborhood search extension and their combination approaches have been proposed, including Simple-swap, 2-Opt, 3-Opt, and 4-Opt. The performance of the proposed mechanism was compared and analyzed using benchmark dataset. The results from experimental work confirmed that the proposed approach provides better accuracy with suitable time.