Exchange Rates Forecasting：An Application of ARIMA Models

• Main Authors: Xiang-Rou Kuan, 管相柔
• Other Authors: Yi-Nung Yang
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/03088242918697259288

碩士 === 中原大學 === 國際貿易研究所 === 95 === Abstract The purpose of this paper to adopt different frequency data（daily and monthly data）. This paper use forecasting models, which is the univariate time series model of Box-Jenkins ARIMA models as a forecasting approach. Using the spot exchange rates between US Dollar against New Taiwan Dollar、Japanese Yen、Euro Dollar as our empirical data, and try to find a reasonable and efficient models for predicting the exchange rates of models. Finally, this paper will employ RMSE、MAE、MAPE as our standard evaluating principles, and to evaluate their forecasting performance in out-of-sample models. And inspected that high frequency whether to compare the low frequency best forecastability. The result indicates that the three kind of measurement tools（RMSE, MAE, and MAPE）. Conclusion, We find that prediction errors of daily data are small relatively to the monthly data. Therefore, we compare in the high frequency data to the low frequency best forecastability. Exclude to obtain data cost, enterprise decision-making or individual investor practice operation in the foreign exchange market, may be to use high frequency data to build forecasting models. if we can effectively control the foreign exchange market of trend, and to reduces the decision-making wrong probability, and furthermore, to achieve the investment、arbitrage、hedge, and we can reduce the risks of the exchange rates.