An Improved Salp Swarm Algorithm Based on Spark for Feature Selection

Abstract

Salp swarm algorithm (SSA) is a population-based optimization technique with excellent performance. However, due to the lack of inertial parameters, this algorithm lacks the ability to find the global search for potential solutions. In this paper, we present an improved SSA algorithm based on spark(Spark-BSSA), which takes advantages of the improved global search ability of BSSA and combines with the spark programming mode. Experimental results demonstrated that as the number of cluster nodes increases, the running time of the algorithm decreases. At the same time, the proposed method is evaluated using the real data sets and compared with the binary gentic algorithm (BGA), the binary particle swarm algorithm (BPSO) and the binary gravity search algorithm (BGSA). The method has also higher classification performance and better stability of classification accuracy. Through experimental data analysis, the ultimate goal is to solve the problems of premature convergence and optimality in the original algorithm.