Review of Metaheuristic Techniques for Feature Selection

Due to the rise of high-dimensional datasets in various sectors, feature selection is one of the most important issues. The primary goal of the feature selection problem is to lower the dimension of the feature set while retaining the performance accuracy. Several techniques of feature selection have been developed to obtain the optimal subset of features. To obtain optimal features, metaheuristics are modern optimization techniques that are used by the research community. In this paper, four groups of metaheuristic techniques have been identified based on their behavior. The classifier name, datasets, and assessment metrics for the metaheuristic methods used to solve the feature selection challenges are provided. After reviewing the papers, difficulties and problems are also observed while trying to use various metaheuristic methods to find the best subset of features. For those researchers who desire to continue their work on creating or refining metaheuristic techniques for feature selection, several research gaps are also mentioned.