Hybrid Metaheuristic Algorithms for Feature Selection in Classification: A Systematic Literature Review

Abstract The effectiveness of a machine learning model is significantly impacted by feature selection. Feature selection is one of the most popular and highly effective techniques for eliminating irrelevant and redundant features to enhance the relevance of the collected data and improve the effectiveness of classification tasks. Feature selection is challenging because of the intricate relationship between features and large search space, which grows exponentially with the number of existing features in the original dataset. Metaheuristic algorithms are the most effective techniques for managing feature selection due to their robustness, intelligibility, and effectiveness in solving intricate optimization problems. Recent studies have focused on using hybrid metaheuristics as feature selection approaches. This systematic literature review explores recent studies from 2019 to 2023 that used hybrid metaheuristic algorithms for feature selection in classification. This paper aims to understand the existing hybrid algorithms, the goal of hybridization, the type of hybridization, and their application domains. Moreover, crucial parameters, fitness and transfer functions, initial population method, feature selection approach, classification algorithm, evaluation criteria, and statistical test are also investigated in this paper. A list of 30 relevant papers in line with the topic were extracted and examined to develop new insights in the domain of feature selection in classification. The focus is on a single fitness function (single objective). However, feature selection can be seen as a multi-objective problem, making hybridization in multi-objective feature selection problems a future research work for scholars.