An improved bees algorithm local search mechanism for numerical dataset

Bees Algorithm (BA), a heuristic optimization procedure, represents one of the fundamental search techniques is based on the food foraging activities of bees. This algorithm performs a kind of exploitative neighbourhoods search combined with random explorative search. However, the main issue of BA is that it requires long computational time as well as numerous computational processes to obtain a good solution, especially in more complicated issues. This approach does not guarantee any
optimum solutions for the problem mainly because of lack of accuracy. To solve this
issue, the local search in the BA is investigated by Simple swap, 2-Opt and 3-Opt were proposed as Massudi methods for Bees Algorithm Feature Selection (BAFS). In this
study, the proposed extension methods is 4-Opt as search neighbourhood is presented. This proposal was implemented and comprehensively compares and analyse their performances with respect to accuracy and time. Furthermore, in this study the feature selection algorithm is implemented and tested using most popular dataset from Machine Learning Repository (UCI). The obtained results from experimental work confirmed that the proposed extension of the search neighbourhood including 4-Opt approach has provided better accuracy with suitable time than the Massudi methods.