Prediction Of Stock Return Rate Based On Grey GARCH-BP Portfolio Model

Prediction of stock return is a hot and difficult issue in the academic and practical circles nowadays.Stock volatility is one of the important characteristics of stock market.The stock return series has obvious characteristics of "peak and heavy tail","volatility agglomeration" and non-linearity.ARCH model and GARCH model are the main methods to study stock return volatility.Neural network model and grey GM model are the main methods to study stock return volatility.It has strong generalization ability,adaptive ability and outstanding prediction ability.Therefore,this paper explores the combination of GARCH model,grey model and BP model to forecast stock returns.This paper takes the stock return of Ping An Bank(000001)as the research object,chooses the daily return data of Ping An Bank(000001)from June 20,2013 to June 15,2018,and uses the methods of grey GM model,grey GARCH model and GARCH-BP combination model to forecast the return of Ping An Bank by using the technology of variance modeling.The results show the prediction effect ratio of GARCH-BP combination model.More satisfactory.Firstly,this paper briefly introduces the research status of grey GM model and the basic theory of grey GM model,divides the daily return data of Ping An Bank into training set and testing set,uses the daily return training set data of Ping An Bank to model,determines the time length of GM model prediction through random sampling and multiple simulation,and establishes GM(1,1)model.GM(1,1)model and metabolic GM(1,1)model are used to predict the return data of training set and test set.The error sequence of GM(1,1)return is obtained.The results of empirical analysis of GM(1,1)model show that there is room for further improvement of the accuracy of the prediction results of test set.Secondly,this paper introduces the research situation of GARCH model and grey GARCH,and further introduces the basic theory and modeling method of grey GARCH.Using the daily return data of Ping An Bank to model,through descriptive analysis,correlation analysis and stationarity test of training set data,the relevant parameters of GARCH model are determined.The residual terms of GARCH model are modified by metabolic GM(1,1)model.The modified residual terms are linearly combined with the residual terms of GARCH model,and the linear combination coefficients are advanced by regression analysis.The grey GARCH model is used to forecast the return rate,and the error sequence of the return rate is obtained.The result of the empirical analysis of the grey GARCH model is better than that of the GM(1,1)model.Thirdly,the research status and modeling theory and method of BP neural network model are introduced.Using the residual sequence of GM(1,1)model and grey GARCH model,the BP model of neural network with 2-X-1 model structure is established.The number of nodes in hidden layer is determined by repeated training of training set.The residual is corrected by BP neural network model.The residual output value of neural network and the predicted value of grey GARCH model are superimposed to obtain the grey GARCH-BP combination.The final prediction value and error value of the model,and the empirical analysis results of the grey GARCH-BP combination model show that the accuracy of the model is better than that of the grey GARCH model.Finally,using RMSE,MAE and MAPE as three evaluation indicators,the prediction effects of gray model,gray GARCH model and gray GARCH-BP model are compared and evaluated.The evaluation results show that the gray GARCH-BP model has the highest prediction accuracy,followed by the gray GARCH model and the gray model is the worst.So far,the grey GARCH-BP portfolio model proposed in this paper can play a better role in predicting stock returns.