A Survey on Metaheuristic Algorithms Utilized for Feature Selection

Document Type : Original Article

Authors

1 Information Systems and Digital Transformation Administration

2 Faculty of Computers and Artificial Intelligence, Helwan University

Abstract

Feature selection (FS) has become an important step in data preprocessing, not only in data mining (DM) but also in machine learning (ML), owing to the ever-increasing amount of data. To tackle the challenge of selecting relevant features, many techniques have been proposed over time. In recent years, metaheuristic algorithms for feature selection, which are divided into swarm intelligence (SI), evolutionary algorithms(EA), and physics base algorithms(PA), have become increasingly popular and are now considered the most effective option compared to other methods. Our research aims to investigate the current challenges associated with feature selection using metaheuristic algorithms. We are particularly interested in exploring the outstanding performance of numerous metaheuristic algorithms for feature selection that have been observed in various areas over the past fifteen years. The study was segmented into several parts. At first, we presented the idea of feature selection. After that, we analyzed the scientific context that elaborated the issues related to feature selection and metaheuristic algorithms. Later on, we investigated the architecture of these algorithms and then proceeded towards the major metaheuristic algorithms that are commonly used in the domain of feature selection. Ultimately, we highlight the primary sources of datasets and some of the machine learning classifiers that are utilized in this field.

Keywords

Main Subjects

Files

Share

How to cite

Statistics