在經濟領域有學者研究匯率的經濟預測模型；在電腦科學領域學者利用了機器學習模型來預測匯率，但是跨領域學者常僅與電腦科學的模型比較，很少比較經濟預測模型與機器學習模型的預測績效，本文利用馬可夫轉換模型(Markov Switching Model)及向量誤差修正模型(Vector Error Correction Model)來與機器學習(Machine Learning)比較預測能力的優劣，結果發現在短期經濟預測模型與機器學習模型並無明顯的差異，而在長期機器學習模型有比較好的預測能力。;In the field of economics, scholars studied how to forecast exchange rates by economic models. In the field of computer science, scholars applied machine learning approach to forecast exchange rates. Although cross-disciplinary scholars often compare their empirical model with computer science models, they hardly compare the performance of economic forecasting models with the performance of machine learning approach. In this thesis, we applied Markov Switching Model, Vector Error Correction Model and Machine Learning approach to forecast the exchange rate of new Taiwan dollar. Besides, we compared the outcome of economic model with the outcome of machine learning models. The results show that, in the short run forecast horizon, there are insignificant difference between the economic models and the machine learning models. In the long run forecast horizon, there are significant differences between economic models and the machine learning models.
