【摘要】 将一种基于特征提取的ε-不灵敏支持向量机方法用于非线性系统辨识.对输入输出数据首先进行核主元特征提取,将特征提取后的数据作为支持向量机的训练数据.将该方法与基于主元特征提取的方法和直接应用ε-不灵敏支持向量机的方法进行含噪和不含噪情况下的仿真比较,结果表明,方法的拟合性能和抗干扰能力优于其他两种方法.更多还原

【Abstract】 A feature extraction based ε-insensitive support vector machine(SVM) method is applied to nonlinear system identification.First,kernel principal component analysis(KPCA) is used to input-output data for feature extraction,which is used as the training data to ε-insensitive SVM.Then,we compare the proposed method with ε-insensitive SVM method and PCA based ε-insensitive SVM.the results show that KPCA based method perform better than ε-insensitive SVM and PCA based ε-insensitive SVM methods.更多还原