Abstract: Support Vector Machine (SVM) is a novel modeling method that is valuable in regression and classification. Kernel parameters setting in the SVM training process, along with the feature selection, significantly affects system identification accuracy. The objective of this study is to obtain the better algorithms for better prediction accuracy. This study develops Principle Component Analysis (PCA) for feature selection and a grid searching and k-fold cross validation (GSCV) approach for parameter optimization in the SVM. Numerical and engineering results indicate that SVM based on PCA can be used for identification of nonlinear functions with related input variables, while the SVM based on GSCV is useful for complex system identification with limited number with kinds of uncertainties.