Binary Black Widow Optimization Approach for Feature Selection

Feature selection is a process of reduction of irrelevant, negligible, noisy features from data sets so as to obtain better performance measurements with fewer features. Throughout the literature, various methods are presented that use different approaches to get through this difficult problem, prevalently. In this study, a binary variant of the Black Widow Optimization (BWO) is proposed in a wrapper mode for the purpose of feature selection. The BWO algorithm has early convergence ability on continuous problems and that characteristic is also effective for finding an optimum solution in feature selection problem. The proposed approach compared with state-of-the-art and widely used approaches such as Binary Particle Swarm Optimization (BPSO and VPSO), Binary Grey Wolf Optimization (BGWO1 and BGWO2). The performance of these algorithms is assessed over 20 benchmark data sets from the UCI repository. The results show that the proposed binary method can be utilized effectively in discrete problems such as feature selection.