Construction and Application of WD-GA-SVR Model - CSI 300 Index Forecast
In order to improve the predictive accuracy of SVM on financial time series, we propose a WD-GA-SVR model in this paper. First, do wavelet denoising (WD) to financial time series to eliminate the interference noise in the original sequence; and use the genetical gorithm (GA) for parameter optimization to make up selecte defects of the parameters c and g; use the obtained optimal parameter values to train to formate the optimal SVM learning machine. The forecast of the order time delay data of CSI 300 index and the comparative analysis of the SVR model, GA-SVR and the WD-GA-SVR model predictions indicate that the WD-GA-SVR model significantly improve the robustness of the prediction model and forecast accuracy, can provide a reference for policy makers and investors.