The Influence of Time Series model Forecasting Accuracy On Wavelet Analysis -Evidence from NTD/USD exchange rate

• Main Authors: Wu, Hsiu Hung, 吳修宏
• Other Authors: Liao, Szu Lang
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/e9y787

碩士 === 國立政治大學 === 金融研究所 === 101 === This paper illustrates an application of wavelets transform method with “singal analysis methods”. The entire procedure can be roughly divided into three steps: wavelet decomposition, signal extension and wavelet reconstruction. In the step of wavelet decomposition, we divide the data into low-signal and high-signal time-series sub data just like the long term trend and the short term volatility in time-series. Second, we apply the ARMA and ARMA-GARCH model to forecast the exchange rate separately and finally reconstruct the two predicting value from the best fitting model to form the forecasting exchange rate which could be compared to the real value. It could be concluded in this study that if we apply the MAE and RMSE index to evaluate the predicting result which is generated from the time-series model with the wavelets transformation of the data beforehand, the forecasting accuracy could be enhanced no matter the data are in daily, weekly or monthly type. In other words, no matter what type of time series data is, the wavelets transform method does enhance the forecasting accuracy.