Research On Stock Index Forecasting Based On GARCH And BP Neural Network Model

With the continuous development of economic and financial markets,stock investment has gradually become people's daily investment method.Predicting the trend of stocks has become an increasingly popular topic.The stock price index can better summarize the stage trend of stocks and reveal the inherent law of stock price changes.Therefore,studying the stock index in the stock market has very important application significance.GARCH model can well describe the heteroscedasticity of financial product risk,but there is a certain lag;BP neural network has a great advantage in the prediction of nonlinear time series,but due to its frequent overfitting phenomenon,it will also reduce the accuracy of forecasting.The two models have a complementary relationship,so this paper selects these two models to form two combinations,compares the two combination forecast results to select a more appropriate combination model to achieve a more accurate forecast of stock index prices.The article compares and analyzes the characteristics of the GARCH model and the BP neural network model,and combines the above models with the establishment of the Shanghai Composite Index,and uses the weight distribution method and the residual correction method to establish a linear combination model based on the mean square error for weight distribution.And the combined model based on residual correction.After empirical analysis based on the above ideas,this article uses mean square error,root mean square error,average absolute error,and average absolute percentage error as evaluation indicators to make short-term,midterm and long-term forecasts on the two combined models,GARCH model and BP neural network model Evaluation.The research results show that the weight distribution method has the smallest forecast error results and the short-term prediction effect is the best,indicating that the combination model based on the weight distribution method is more effective in forecasting and can achieve more accurate stock price index forecasts in the short term,which has certain implications for investors Realistic meaning.