Metaheuristic Techniques in Feature Selection: A Concise Review

Abstract

Feature selection is an essential process in machine learning, designed to diminish the dimensionality of the feature set while preserving performance accuracy. Since the 1970s, various approaches for feature selection have been proposed, with metaheuristic algorithms being the most effective. This survey analyzes the prominent metaheuristic feature selection algorithms, emphasizing their efficacy in exploration/exploitation operators, selection methodologies, transfer functions, fitness evaluations, and parameter optimization strategies. The paper provides a comprehensive literature analysis on addressing feature selection issues with metaheuristic algorithms. Metaheuristic algorithms are categorized into four types according to their behavior, with a compilation of more than one hundred listed. The paper addresses obstacles and issues in acquiring the optimal feature subset through various metaheuristic algorithms and identifies research gaps for researcher.