Feature Selection Using Chaotic Salp Swarm Algorithm for Data Classification

Abstract

Salp swarm algorithm (SSA) is a recently created bio-inspired optimization algorithm presented in 2017 which is based on the swarming mechanism of salps. Despite high performance of SSA, slow convergence speed and getting stuck in local optima are two disadvantages of SSA. This paper introduces a novel chaotic SSA algorithm (CSSA) to avoid these weaknesses, where chaotic maps are used to enhance the performance of SSA algorithm. The CSSA algorithm is incorporated with the K-nearest neighbor classifier to solve the feature selection problem, in which twenty-seven datasets are used to assess the performance of CSSA algorithm. The results confirmed that the proposed chaotic SSA (especially Tent map) produced superior results compared to standard SSA and other optimization algorithms.