Prediction Of CSI300 Index Based On Morlet Wavelet Kernel Function Support Vector Machine

The stock market is a complex system which is influenced by many factors,and the stock price fluctuations tend to be random.Therefore,it is very difficult to forecast the price.This also led to a large number of scholars in and abroad to study the prediction of stock price,putting forward a variety of stock price forecasting models.Among these models,support vector machine(SVM)based on statistical learning theory is a good prediction method.Wavelet analysis is a commonly used signal analysis method,which has the characteristics of multi-scale and localization.The wavelet analysis could be introduced into the support vector machine(SVM)model,and training SVM regression model based on wavelet kernel function is an effective method to improve the ability of SVM to predict the non-stationary time series.The stock price series belongs to the typical non-stationary time series,so the SVM regression model based on wavelet kernel function can be applied to the prediction of stock price,improving the theoretical framework of stock price forecasting.On the basis of summarizing the existing theories of stock price forecasting,combining wavelet analysis theory and support vector machine theory,Morlet wavelet function is used as the kernel function of SVM model,and train the SVM regression model based on Morlet wavelet kernel function.In order to verify the model could be applied to forecast the stock price,this paper takes CSI300 index as the research object,using the SVM regression model based on Morlet wavelet kernel function and other three SVM regression models based on traditional kernel functions to predict CSI300 index.Genetic algorithm and particle swarm optimization could be used to determine the optimal parameters of these models.By comparing the prediction results of each model,it can be concluded that the prediction performance of the SVM regression model based on Morlet wavelet kernel function optimized by particle swarm optimization is the best.It proves that the SVM regression model based on Morlet wavelet kernel function can be successfully applied to the prediction of financial time series.Finally,the BP neural network model is trained,and forecast the CSI300 index with the model.The prediction results show that the prediction accuracy of SVM regression model based on Morlet wavelet kernel function is better than the prediction accuracy of BP neural network model,which proves that in finite samples,the SVM regression model based on Morlet wavelet kernel function is more effective as a predictive model.