A Novel Chaotic Interior Search Algorithm for Global Optimization and Feature Selection

Interior Search Algorithm (ISA) is a recently proposed metaheuristic inspired by the beautification of objects and mirrors. However, similar to most of the metaheuristic algorithms, ISA also encounters two problems, i.e., entrapment in local optima and slow convergence speed. In the past, chaos theory has been successfully employed to solve such problems. In this study, 10 chaotic maps are embedded to improve the convergence rate as well as the resulting accuracy of the ISA algorithms. The proposed Chaotic Interior Search Algorithm (CISA) is validated on a diverse subset of 13 benchmark functions having unimodal and multimodal properties. The simulation results demonstrate that the chaotic maps (especially tent map) are able to significantly boost the performance of ISA. Furthermore, CISA is employed as a feature selection technique in which the aim is to remove features which may comprise irrelevant or redundant information in order to minimize the classification error rate. The performance of the proposed approaches is compared with five state-of-the-art algorithms over 21 data sets and the results proved the potential of the proposed binary approaches in searching the optimal feature subsets.