本研究探討深度學習 LSTM 模型與線性迴歸 OLS 模型對於新台幣兌美元匯 率走勢預測表現，根據未拋補利率平價模型(UIRP)、購買力平價模型(PPP)、 貨幣模型(MF)以及泰勒模型(Taylor)選擇總體經濟變數，並且將總體經濟變 數區分為 Decouple 與 Couple 型態納入 LSTM 模型與 OLS 模型進行預測，最後 以 R square、Theil 比率作為衡量預測能力標準，除此之外，本研究進一步比較各 模型的方向預測表現與交易策略表現，分別利用方向準確率與夏普比率作為衡量 準則。

實證結果顯示，LSTM 模型在匯率預測能力、方向準確率以及交易策略表現 皆優於 OLS 模型，其中以 Recursive LSTM 模型表現最佳。在總體經濟變數方面， MF 整體表現較 UIRP、PPP 以及 Taylor 差，UIRP、PPP 以及 Taylor 表現依據總 經變數 Couple 型態與 Decouple 型態而有些微不同，Couple 型態下 3 種經濟變數 整體表現不相上下，而 Decouple 型態下 UIRP 整體表現優於其他 3 種經濟變數 組合。

This paper explores the performance of deep learning LSTM model and linear regression OLS model for the prediction of the exchange rate between NT dollars and US dollars. I select the economic variables according to the uncovered interest rate parity model, the purchasing power parity model (PPP), the currency fundamental model (MF) and the Taylor rule model (Taylor), and divide all economic variables into Decouple and Couple types for prediction. R square and Theil ratio are used as the standard to measure the prediction ability. In addition, this paper also compares the hedging performance and economic benefit of the model which are measured by the hedging accuracy rate and the Sharpe ratio, respectively.

The results show that the LSTM model outperforms the OLS model in exchange rate prediction ability, hedging accuracy and economic benefit. The Recursive LSTM model performs the best. In the economic variables, the overall performance of MF is worse than that of UIRP, PPP and Taylor model. The performance of UIRP, PPP and Taylor is different according to the Couple type and Decouple type. The UIRP, PPP and Taylor model under the Couple type get similar performance. The comprehensive performance of UIRP under the Decouple type is better than that of the other three economic variable combinations.

[1] 程智男、林建秀、尤保傑(2016)。有效匯率預測模型與避險績效比較。應 用經濟論叢，99，37-82。
[2] Amat, Christophe, Tomasz Michalski, and Gilles Stoltz (2018). Fundamentals and exchange rate forecasta- bility with simple machine learning methods. Journal of International Money and Finance, 88: 1–24.
[3] Cheung, Y. W., M. D. Chinn, and A. G. Pascual, (2005). Empirical Exchange Rate Models of the Nineties: Are Any Fit to Survive? Journal of International Money and Finance, 24: 1150-1175.
[4] Della Corte, P. and I. Tsiakas, (2011). Statistical and Economic Methods for Evaluating Exchange Rate Predictability. in James, J., I. Marsh, and L. Sarno, ed., Handbooks of Exchange Rates, 239-283, NJ: Wiley and Sons Inc. Press.
[5] Engel, C. and K. D. West, (2005). Exchange Rates and Fundamentals, Journal of Political Economy. 113: 485-517.
[6] Engel, C. and K. D. West, (2006). Taylor Rules and the Deutschmark-dollar Real Exchange Rate. Journal of Money, Credit and Banking, 38: 1175-1994.
[7] Engel, C., N. C. Mark, and K. D. West, (2007). Exchange Rate Models are Not as Bad as You Think. NBER Working Paper, No. 13318.
[8] Groen, J. J. J., (2000). The Monetary Exchange Rate Model as a Long-run Phenomenon. Journal of International Economics, 52: 299-319.
[9] Isha S Meshram, Prajakta J Kulal (2021). A comparative study of SVM, LSTM and LR algorithms for stock market prediction using OHLS data. International Research Journal of Modernization in Engineering Technology and Science, 3: 1316-1322.
[10] Kumar, P.H.; Patil, S.B. (2018). Forecasting volatility trend of INR USD currency pair with deep learning LSTM techniques. In Proceedings of the 3rd International Conference on Computational Systems and Information Technology for Sustainable Solutions (CSITSS), Bengaluru, India, 20–22 December 2018; pp. 91–97.
42
[11] Mark, N. C. and D. Sul, (2001). Nominal Exchange Rates and Monetary Fundamentals: Evidence from a Small Post-bretton Woods Panel. Journal of International Economics, 53: 29-52.
[12] Mark, N. C., (1995). Exchange Rates and Fundamentals: Evidence on Long- horizon Predictability. American Economic Review, 85: 201-218.
[13] Mark, N. C., (2009). Changing Monetary Policy Rules, Learning, and Real Exchange Rate Dynamics. Journal of Money, Credit and Banking, 41: 1047-1070.
[14] Meese, R. A. and K. Rogoff, (1983). Empirical Exchange Rate Models of the Seventies: Do They Fit Out of Sample? Journal of International Economics, 14: 3- 24.
[15] Molodtsova, T. and D. H. Papell, (2009). Out-of-Sample Exchange Rate Predictability with Taylor Rule Fundamentals? Journal of International Economics, 77: 167-180.
[16] Molodtsova, T., and D. H. Papell, (2012). Taylor Rule Exchange Rate Forecasting During the Financial Crisis. NBER Working Paper, No. 18330.
[17] Nelson,M.Q., Pereira, A.C.M. and deOliveira,R.A. (2017). Stockmarketsprice movement prediction with LSTM neural networks. International Joint Conference on Neural Networks (IJCNN). Piscataway, NJ: IEEE, pp. 1419-1426.
[18] Qi, L., Khushi, M., Poon, J. (2020). Event-driven LSTM for forex price prediction. In Proceedings of the 2020 IEEE Asia-Pacific Conference on Computer Science and Data Engineering (CSDE), Gold Coast, Australia, pp. 16-18.
[19] Taylor, A. M. and M. P. Taylor, (2004). The Purchasing Power Parity Debate. Journal of Economic Perspectives, 18: 135-158.
[20] Taylor, J. B., (1993). Discretion versus Policy Rules in Practice. Carnegie- Rochester Conference Series on Public Policy, 39: 195-214.
[21] Yaxin Qu and Xue Zhao (2019). Application of LSTM neural network in forecasting foreign exchange price. Journal of Physics: Conference Series.
[22] Yildirim, D. C., Toroslu, I. H., & Fiore, U. (2021). Forecasting directional movement of Forex data using LSTM with technical and macroeconomic indicators. Financial Innovation, 7(1), 1–36.