| Summary: | 碩士 === 國立成功大學 === 會計學系 === 102 === The exchange rate is time series data that unstable, complex and difficult to predict. In tradition, the forecasting in time series data is to use statistical method. Generally speaking, autoregressive integrated moving-average (ARIMA) model for forecasting in linear data is quite good. Hence, we use the sample data to establish the ARIMA model at first and derive the linear predictive values. The mathematical financial model, Cox-Ingersoll-Ross (CIR) model also be used to predict the exchange rate through the uncover interest rate parity (UIRP). Therefore, second, we use STRIPS bonds of U.S. and Japan to obtain the estimated CIR models to predict the exchange rate in our sample period, Jan.2, 2012 to Mar. 30, 2012. We use the moving window method to generate the estimated exchange rates. Finally, in order to measure the predictive power, we calculate the root mean square error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) of the forecasting models. The empirical results show that the predictive power of the CIR model is significantly better than traditional ARIMA model.
|
|---|
