Parameter optimization of support vector machine based on improved sparrow search algorithm

Abstract

Support vector machine is widely used in various fields because of its excellent generalization performance. However, the selection of its parameters directly affects the accuracy of the final results. An improved sparrow search algorithm (ISSA) is proposed to optimize the parameters of support vector machines. The ISSA algorithm improves the original algorithm from three aspects: replacing random method with optimal point set initialization population, changing the explorer position update formula, and adopting adaptive mutation mechanism. The UCI standard data set was selected to compare the SVM optimized by ISSA algorithm with the original SVM, the SVM optimized by the genetic algorithm, the particle swarm optimization algorithm and the basic sparrow search algorithm, respectively. The experimental results show that the classification accuracy of the SVM optimized by ISSA algorithm is significantly improved, and the generalization performance is further improved.