Abstract:As an effective tool in pattern recognition and machine learning, support vector machine (SVM) has been adopted abroad. In developing a successful SVM classifier, eliminating noise and extracting feature are very important. This paper proposes the application of kernel PCA to SVM for feature extraction. Then PSO Algorithm is adopted to optimization of these parameters in SVM. The novel time series analysis model integrates the advantage of wavelet, PSO, KPCA and SVM. Compared with other predictors, this model has greater generality ability and higher accuracy.

The Combining Kernel Principal Component Analysis with Support Vector Machines for Time Series Prediction Model