改进粒子群算法优化的支持向量机及其应用

A support vector machine based on an improved particle swarm optimization algorithm and its application

传统粒子群优化（particle swarm optimization,PSO）算法主要包含两方面问题,即易陷入局部极小和后期震荡严重,为此引入混沌序列来初始化粒子群的位置,并在简化的粒子群数学模型上从两个方面对其进行了改进。本文利用改进的PSO算法对支持向量机（support vector machine,SVM）的参数进行优化,仿真实验结果表明：与SVM、PSO-SVM以及遗传算法（genetic algorithm,GA）优化的SVM（GA-SVM）相比,改进PSO优化的SVM（IPSO-SVM）算法具有较高的分类准确率,并且与PSO-SVM算法相比,准确率提高了3%~5%,与PSO-SVM算法以及GA-SVM算法相比,IPSO-SVM的训练和泛化速度都明显提高。本文将IPSO-SVM算法应用到遥感影像的分类中,分类结果表明,与PSO-SVM算法相比IPSO-SVM算法具有更好的分类结果。

The traditional particle swarm optimization （PSO） algorithm has two types of defects, i.e., easily falling into the local minima and severe shock in the later stages. To address these problems, this paper proposes an im-proved PSO （ IPSO） algorithm, which introduces a chaotic sequence to initialize the positions of the particle swarm, and optimizes it from two aspects based on a simplified PSO mathematical model. This paper used IPSO to optimize the model parameters of a support vector machine （SVM）. The simulation experiment results proved that compared with SVM, PSO-SVM, and genetic algorithm optimized SVM （GA-SVM）, the improved PSO optimized SVM（IPSO-SVM） has higher classification accuracy. Compared with PSO-SVM, IPSO-SVM increased the accuracyby 3%～5%, and compared with PSO-SVM and GA-SVM, the training and testing time was obviously reduced. Fi-nally, the IPSO-SVM algorithm was applied to the classification of remote sensing images; the results proved that IPSO-SVM offers better classification solutions than PSO-SVM.