Forecasting RMB Exchange Rate With A Hybrid ARIMA And GRNN Model

With the high development of China's economy and the accelerating process of exchange rate marketization,the RMB exchange rate plays a more and more important role in the domestic and foreign economy.From August 11,2015,to promote marketization of RMB exchange rate and let RMB join the SDR smoothly,PBC has abandoned the intervention to the daily exchange rate of RMB against the US dollars,the wider exchange rate fluctuates,the stronger uncertainty short-term exchange rate movement has.Under this background,enhancement RBM exchange rate volatility forecast is of great theoretical significance and practical value to reduce the production cost and avoiding exchange rate risk for Chinese enterprises.Based on researches of the domestic and foreign scholars for exchange rate,according to the composite characteristics of exchange rate time series with linear and nonlinear,and to take advantage of superiority about ARIMA and GRNN model in linear and nonlinear space,we establish a hybrid ARIMA and GRNN model to forecast and analysis volatility of the RMB against the US dollar time series.There are five parts in this paper.The first part introduces the background and the significance of the research on exchange rate and reviews the prediction techniques of exchange rate of scholars at home and abroad.The main content and the research method of this paper are proposed.The second part reviews the development of exchange rate system of RMB and the developmental tendency of the exchange rate of RMB against the US dollar under different systems.In addition,the influence of the fluctuation in exchange rate on our daily life and national economy is analyzed.In the third part,the relevant theories and the modeling procedures of ARIMA as well as the theories and parameter setting of GRNN are introduced and the rationale and modeling procedure of combined model of ARIMA and GRNN are illustrated.The fourth part predicts and analyzes the time series of RMB against the US dollar exchange reference rate.Firstly,we use ARIMA model to forecast the time series of daily RMB against the US dollar middle exchange reference rate data to acquire the linear main part.Secondly,we use GRNN model to forecast the nonlinear residual of ARIMA model.Finally,the forecasting result of the time series of RMB against the US dollar exchange reference rate is compounded.The forecasting result shows that a hybrid ARIMA and GRNN model has a better forecasting effect than ARIMA model or GRNN model and that GRNN model has a better predictive effect than ARIMA model.Based on the analysis and the summary of the experimental results,the fifth part also points out the limitations of the present study and the suggestions for future study.