Chaotic binary reptile search algorithm and its feature selection applications

Feature selection (FS) is known as the most challenging problem in the Machine Learning field. FS can be considered an optimization problem that requires an efficient method to prepare its optimal subset of relative features. This article introduces a new FS method-based wrapper scheme that mixes chaotic maps (CMs) and binary Reptile Search Algorithm (RSA) called CRSA, employed to address various FS problems. In this method, different chaotic maps are included with the main ideas of the RSA algorithm. The objective function is revealed to combine three objectives: maximizing the classification accuracy, the number of chosen features, and the complexity of produced wrapper models. To assess the achievement of the proposed methods, 20 UCI datasets are applied, and the results are compared with other well-known methods. The results showed the superiority of the introduced method in bettering other well-known techniques, particularly when applying binary RSA with Tent CM.