Comprehensive Learning Strategy Enhanced Chaotic Whale Optimization for High-dimensional Feature Selection

Abstract

Feature selection (FS) is an adequate data pre-processing method that reduces the dimensionality of datasets and is used in bioinformatics, finance, and medicine. Traditional FS approaches, however, frequently struggle to identify the most important characteristics when dealing with high-dimensional information. To alleviate the imbalance of explore search ability and exploit search ability of the Whale Optimization Algorithm (WOA), we propose an enhanced WOA, namely SCLWOA, that incorporates sine chaos and comprehensive learning (CL) strategies. Among them, the CL mechanism contributes to improving the ability to explore. At the same time, the sine chaos is used to enhance the exploitation capacity and help the optimizer to gain a better initial solution. The hybrid performance of SCLWOA was evaluated comprehensively on IEEE CEC2017 test functions, including its qualitative analysis and comparisons with other optimizers. The results demonstrate that SCLWOA is superior to other algorithms in accuracy and converges faster than others. Besides, the variant of Binary SCLWOA (BSCLWOA) and other binary optimizers obtained by the mapping function was evaluated on 12 UCI data sets. Subsequently, BSCLWOA has proven very competitive in classification precision and feature reduction.