An accelerated sine mapping whale optimizer for feature selection

Helong Yu,Zisong Zhao,Ali Asghar Heidari,Li Ma,Monia Hamdi,Romany F Mansour,Huiling Chen

DOI: https://doi.org/10.1016/j.isci.2023.107896

IF: 5.8

2023-09-14

iScience

Abstract:An improved whale optimization algorithm (SWEWOA) is presented for global optimization issues. Firstly, the sine mapping initialization strategy (SS) is used to generate the population. Secondly, the escape energy (EE) is introduced to balance the exploration and exploitation of WOA. Finally, the wormhole search (WS) strengthens the capacity for exploitation. The hybrid design effectively reinforces the optimization capability of SWEWOA. To prove the effectiveness of the design, SWEWOA is performed in two test sets, CEC 2017 and 2022, respectively. The advantage of SWEWOA is demonstrated in 26 superior comparison algorithms. Then a new feature selection method called BSWEWOA-KELM is developed based on the binary SWEWOA and kernel extreme learning machine (KELM). To verify its performance, 8 high-performance algorithms are selected and experimentally studied in 16 public datasets of different difficulty. The test results demonstrate that SWEWOA performs excellently in selecting the most valuable features for classification problems.