An Empirical Study Of RMB Exchange Rate Fluctuation Based On Nonlinear GARCH Model

With the acceleration improvement of the integration of global economy and international financial market,the international market continuously improved and developed.Following the increasing dependence on trade among countries and the closer financial system cooperation,the international monetary market becomes more marketizational and linkage.However,since this year,the United States has gradually tended to trade unilateralism.It constantly used anti-dumping and other reasons to provoke economic and trade frictions against our country,which makes people worried about the prospects of Sino-US economic and trade relations.Exchange rate,as a bridge connecting economic and trade exchanges between the two countries,not only serves as a barometer reflecting of the current economic situation of the two countries,but also directly affects the import and export of their countries.Therefore,in order to prevent the United States from further expanding economic and trade frictions,or let the exchange rate is used as a direct weapon to affect the economic development of our country,we should temper our own currency.At the same time,we should put the function of exchange rate,monitoring and forecasting,at the core of future economic regulation and control.However,the reality is that,compared with the current status of exchange rate in China’s economy,academic research and exploration on exchange rate forecasting is still in a groping state.Using empirical and quantitative research methods,this paper tests the stationarity of 1708 U.S.dollar-RMB exchange rate samples from January 1,2013 to May 31,2019.According to the stationarity test results of the data,the time series data are transformed into profitability series.On the basis of the data of the profitability series,the first step is to expand to determine the order of the mean equation and construct the parameters of the ARMA model.After Q-statistical correlation graph test,residual ARCH-LM test and residual square correlation graph test,the order structure of ARMA is determined.On this basis,the order of EGARCH-M model and TGARCH-M model is determined.Referring to the rule of AIC and SBIC,the ARMA-EGARCH-M model and ARMA-TGARCH-M model are obtained and completed.The residual test was carried out.In order to realize the non-linearization of GARCH family models,the BP neural network model is used to train and forecast the data of the two models.1685 data from January 1,2013 to April 30,2019 are used as training samples;23 data from May 1,2019 to May 31,2019 are used as test samples and input into the neural network model.The training is carried out,and the fitting results are evaluatedaccording to the magnitude of the error of the output results and the unequal coefficients of RMSE and Theil and the magnitude of MAE.Suggestions and conclusions on the exchange rate fluctuation itself and the prediction model are put forward.