Stock Price Prediction Based On PSO-GARCH-SVR Model With High Frequency Data

The situation of each sector and industry in China's stock market is different.Information asymmetry,policies,and the international financial environment will all affect the stock market.Stock price forecasting has always been a hot topic in the field of statistics.Many scholars are committed to creating new models for predicting stock prices.At present,domestic research on the use of machine learning and deep learning to predict stock prices has become more and more in-depth.In order to improve the accuracy of stock price prediction,this paper takes the per minute high-frequency data of the motherboard and small and medium-sized board market as the research object to improve the prediction model based on SVR.In this paper,the technical analysis index is used as the original variable,and the cross-validation method is used to select the kernel function and parameters for the stock.On this basis,the GARCH model is established to calculate the volatility of the stock price logarithmic yield,and the GARCH-SVR model is established.Then the particle swarm optimization algorithm is used.A cyclic process of continuously updating the weights of input variables is established to establish a PSO-SVR model.Finally,the three models are combined to form a PSO-GARCH-SVR model.After that,this paper selects the 5-minute high-frequency data of the Hushen 300 Index and the SME board index as samples to verify the applicability of the model.This paper uses RMSE and ups and downs prediction accuracy as the evaluation index.The empirical results show that the GARCH-SVR model in the two markets has no obvious optimization effect,and the prediction effect of the PSO-SVR model relative to the SVR model is significantly improved.The PSO-GARCH-SVR model is further optimized based on the PSO-SVR model and has applicability in both markets.This model can provide much reference value for the research in the field of stock price forecasting.