A Modified White Shark Optimizer for Gene Selection Optimization Problem

This paper investigates the application of the White Shark Optimizer (WSO) to gene selection, a fundamental task in machine learning to analyze complex, high-dimensional data such as gene expressions derived from DNA microarray studies. The problem of gene selection is complicated by the presence of redundant and misleading genes that can reduce the effectiveness of machine learning models. We divided gene selection techniques into two categories: candidate-based and envelope-based, focusing on prominent methods such as Kullback-Leibler and Chi-square to measure the difference between the probability distributions. We focus on the novel WSO, which draws inspiration from the predatory strategies of white sharks, famous for their adaptability and efficiency in complex scenarios. The paper presents a modified version of WSO, tailored to the binary nature of gene selection, and evaluates its effectiveness against particle swarm optimization. The results highlight the superior accuracy and robustness of WSO, where the proposed WSO outperformed the compared method in obtaining the accuracy and fitness values of the three datasets.