Abstract
The performance of classification algorithms is highly sensitive to the data dimensionality. High dimensionality may cause many problems to a classifier like overfitting and high computational time. Feature selection (FS) is a key solution to both problems. It aims to reduce the number of features by removing the irrelevant, redundant and noisy data, while trying to keep an acceptable classification accuracy. FS can be formulated as an optimization problem. Metaheuristic algorithms have shown superior performance in solving this type of problems. In this work, a chaotic version of Salp Swarm Algorithm (SSA) is proposed, which is considered one of the recent metaheuristic algorithms. The proposed approach is applied for the first time on feature selection problems. Four different chaotic maps are used to control the balance between the exploration and exploitation in the proposed approach. The proposed approaches are evaluated using twelve real datasets. The comparative results shows that the chaotic maps significantly enhances the performance of the SSA algorithm and outperforms other similar approaches in the literature.