A Comprehensive Analysis of Nature-Inspired Meta-Heuristic Techniques for Feature Selection Problem

• Published in: Archives of computational methods in engineering Vol. 28; no. 3; pp. 1103 - 1127
• Main Authors: Sharma, Manik, Kaur, Prableen
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.05.2021; Springer Nature B.V
• Subjects: Algorithms; Engineering; Feature selection; Heuristic; Heuristic methods; Mathematical and Computational Engineering; Multidisciplinary research; Optimization; Optimization techniques; Original Paper; Redundancy
• ISSN: 1134-3060; 1886-1784
• DOI: 10.1007/s11831-020-09412-6

Meta-heuristics are problem-independent optimization techniques which provide an optimal solution by exploring and exploiting the entire search space iteratively. These techniques have been successfully engaged to solve distinct real-life and multidisciplinary problems. A good amount of literature has been already published on the design and role of various meta-heuristic algorithms and on their variants. The aim of this study is to present a comprehensive analysis of nature-inspired meta-heuristic utilized in the domain of feature selection. A systematic review methodology has been used for synthesis and analysis of one hundered and seventy six articles. It is one of the important multidisciplinary research areas that assist in finding an optimal set of features so that a better rate of classification can be achieved. The concept of feature selection process along with relevance and redundancy metric is briefly elucidated. A categorical list of nature-inspired meta-heuristic techniques has been presented. The major applications of these techniques are explored to highlight the least and most explored areas. The area of disease diagnosis has been extensively assessed. In addition, the special attention has been given on highlighting the role and performance of binary and chaotic variants of different nature-inspired meta-heuristic techniques. The summary of nature-inspired meta-heuristic methods and their variants along with datasets, performance (mean, best, worst, error rate and standard deviation) is also depicted. In addition, the detailed publication trend of meta-heuristic feature selection approaches has also been presented. The research gaps have been identified for the researcher who inclines to design or analyze the performance of divergent meta-heuristic techniques in solving feature selection problem.