基于核函数主元分析的SVM建模方法及应用

SVM Modeling and Application Based on Kernel Function Principal Component Analysis

摘要: 为有效克服线性建模方法在非线性建模方面的不足,将核函数思想引入到主元分析方法(PCA)中,有效提取实验数据中的非线性特征信息,并将其作为支持向量机(SVM)的输入变量,建立工业过程软测量模型。该方法应用于丙烯腈聚合过程中转化率的预报,结果表明:该方法的预测精度优于PCA-SVM方法和KPCA-NN方法。

Abstract: Kernel function is introduced into PCA method to obtain nonlinear character information from experimental data so as to overcome the disadvantage of traditional methods in nonlinear modeling.SVM is utilized in developing soft sensor that uses the nonlinear characteristics of data as the input of SVM.Application shows that the proposed method is effective and superior to both PCA-SVM and KPCA-NN methods in the application of acrylonitrile transforming prediction.