Volatility Model And Prediction Of Internet Monetary Fund Returns Based On Neural Network

The risk of financial assets is usually represented by the uncertainty of returns.Based on the forecast of the yield of Internet money funds,the risk characteristics of internet money funds can be further analyzed to promote the stable development of funds.This paper aims to Internet Monetary Fund's yields are analyzed and forecasted to improve the accuracy of risk analysis.Firstly,GARCH model is one of the most commonly used forecasting models in traditional time series methods.It can handle linear problems very well.This paper first establishes a forecast model of internet money fund return rate based on GARCH model,and analyzes GARCH model's advantages and disadvantages of the yield,due to the unique risks of the Internet money funds,the model's capture of extreme values is not ideal.Secondly,a profit forecasting model based on BP neural network is established.Using the nonlinear characteristics of BP neural network,a better prediction of profitability is achieved.Combined with the linear processing capability of the GARCH model and the nonlinear processing effect of the BP neural network,a yield prediction model based on the BP-GARCH model is established,and the model significantly improved the ability to capture extreme values.Finally,in order to improve the generalization ability of the BP neural network model,the AdaBoost algorithm is used to optimize,and the complementary combination of multiple BP neural networks is realized,and the weight of the BP neural network with better prediction effect is increased,so that the prediction capability is improved.Mean absolute error(MAE),mean square error(MSE),mean error(ME),and orientation accuracy(DA)are used as benchmarks to compare the prediction accuracy of the model.Analyze and compare the pros and cons of each model's forecast for the return rate of internet money funds.The results show that the combination model based on BP-GARCH and the BP neural network model optimized by AdaBoost algorithm fit better to the fluctuation of yield and has higher accuracy.