A Comprehensive Survey on Meta-Heuristic Algorithms for Feature Selection in High-Dimensional Data: Challenges, Applications, and Future Directions

Abstract

Feature selection is crucial for improving machine learning models by reducing dimensionality and lowering computational costs. This survey paper provides an in-depth review on recent advancements in feature selection methods, with a focus on meta-heuristic algorithms. Known for their robustness and effectiveness, these algorithms have become key in tackling the combinatorial challenges in feature selection. The paper categorizes and evaluates various meta-heuristic approaches, including evolutionary algorithms, swarm intelligence, and hybrid techniques, highlighting their strengths, limitations, and applications across different fields. It also examines the integration of meta-heuristics with other optimization methods and machine learning frameworks, identifying current trends and challenges. The paper concludes by discussing future research directions, emphasizing the potential of meta-heuristic-based feature selection in handling high-dimensional data and complex real-world problems.