The Empirical Analysis Of The Volatility Of The Stock Price Based On The Hybrid Model Integrating LSTM With Multiple GARCH Models

Financial market has strong liquidity and certain amplitude of volatility.In real life,it is always a "double-edged sword",on the one hand,it can play an obvious and effective role as a catalyst for the development of social economy and finance,on the other hand,The market itself also has some risks.In the past ten years,Chinaundefineds stock market has experienced two horrifying "stock disaster" large fluctuations.Therefore,it is increasingly urgent to study the volatility of Chinaundefineds stock market.It is very important for us to make rational investment to grasp and predict the volatility of Chinese stock market.In recent years,neural network models and time series models have developed rapidly,and have been used in more and more fields such as financial markets.In this paper,the hybrid model of neural network model and time series model is proposed,and apply it to conduct in-depth research on the mastery and prediction of China's stock market volatility.In this paper,the daily return rate of China Securities 500 Index is taken as the resEGARCH object,and a series of statistical analysis of the selected resEGARCH data is carried out by using the tool software.The logarithmic yield series has the characteristics of fluctuation aggregation,sharp peak and thick tail,right deviation and so on.The sequence is stationary,has no autocorrelation and the residual sequence has ARCH effect.Then establish a GARCH family model based on t distribution and GED distribution.It is found that the fitting effect of GARCH family model under t distribution is optimal and the estimated parameters of each model are obtained and used as the input of mixed LSTM model.Finally,this paper innovatively establishes the hybrid model of LSTM depth neural network model and multiple GARCH model,and forecasts the closing price.The prediction results show that the multi-hybrid model GET-LSTM model fits best in all the mixed LSTM models.The prediction error is minimum.According to the analysis results,some suggestions are put forward from the aspects of information disclosure,government intervention,new variety development,investor education,market supervision and so on.