PSO-SVM模型的构建与应用

为了使支持向量机(SVM)获得更好的分类效果,针对人为选择参数的随机性,提出了利用粒子群算法(PSO)进行参数自动选取的优化方法,构建了PSO-SVM模型.在个人信用评估中,通过对粒子适应度函数的设置来控制造成较大损失的第二类误判,应用结果表明:模型在训练和测试样本中的分类精度可以达到95%,第二类误判率分别仅为0.78%和2.02%.利用PSO对SVM中的参数进行优化,可以避免人为选择的随机性,并且在解决分类问题中表现出较好的稳健性.

Construction and application of PSO-SVM model

Aiming at the randomness of parameter selection,a PSO-SVM model is constructed by using particle swarm optimization(PSO) to achieve higher classification accuracy in support vector machine(SVM).In personal credit scoring,the particles’ fitness function is set up to control the type II error rate.The application results indicate that PSO-SVM model gets high classification accuracy of 95% with low type II error rate of 0.78% and 2.02%.The parameters in SVM are optimized by PSO,so that randomness of parameter selection is avoided and the model shows strong robustness in classification problems.