自布列敦森林協定瓦解後，各國紛紛改採自由浮動的匯率制度。隨著國際間貿易與資金流動日益頻繁，匯率市場充斥著不確定性，其對個人、企業、政府皆具一定匯差損失風險。因此如果能藉由人工智慧的方法進行有效預測，則能有利於制定交易策略。
然而匯率預測長期以來在經濟領域都是困難的問題，在現實世界中有許多因素會導致其匯率價格產生波動，其資料具有的高雜訊、非穩態、非線性等特徵。因此，採用非線性模型進行預測較為合理。而為了提高模型準確性，借用通訊領域經常使用的時頻分析方式，將時間序列視為訊號並分解成不同頻率的分量進行預處裡，以協助之後預測模型進行學習。將拆解成不同分量的本質模態函數分別進行長短期記憶之注意力模型(LSTM-Attention)訓練。另外由於超參數對於預測的準確度有關鍵性的影響，故使用差分進化演算法調整深度學習模型的超參數，以利於實驗的結果。
本研究提出一種LSTM-Attention結合經驗模態分解的模型並且與傳統時間序列整合移動平均自回歸、循環神經網絡相關的門控循環單元、LSTM等模型的效能進行比較，經實驗結果發現，在台幣兌韓圓的資料中，LSTM-Attention(with EMD)模型的表現，在兩項評估指標RMSE與MAPE皆優於其他比較模型。

Since the Bretton Woods agreement collapse, many countries have adopted the a floating exchange rate system. With the international trade and capital flows between countries have become increasingly frequent, the exchange rate market is full of uncertainties and it has a certain degree of risk of exchange rate loss to individuals, enterprises and governments. So it is beneficial to develop trading strategies if we can be effectively predicted by artificial intelligence.
However, exchange rate forecasting has always been a puzzle in the field of Economics, in the real world, there are many factors that cause the exchange rate prices to fluctuate, financial data are inherently noise, non-stationary and non-linearity by default. Therefore, it is reasonable to use nonlinear model to make predictions. In order to improve the accuracy of the model, we use the time- frequency analysis to decompose signals into components of different frequencies for better learning. The long short-term memory with attention(LSTM-Attention) trains the intrinsic mode functions of different portions. Furthermore, the performance of a model highly depends on the hyper-parameters selection. For this reason, a metaheuristic algorithm named Differential Evolution is utilized to identify the suitable hyper-parameters of the model.
In this study, combining empirical modal decomposition and LSTM-Attention model is proposed and compared with the effectiveness of ARIMA、GRU and LSTM. The experimental results show that the performance of the LSTM-Attention (with EMD) model is better than other comparison models in both RMSE and MAPE.

[1] V. Vapnik. (1995). The nature of statistical learning theory. Springer Science
& Business Media.
[2] S. Hochreiter & J. Schmidhuber. (1997). Long short-term memory. Neural
Computation, 9(8), 1735-1780.
[3] E. F. Fama. (1970). Efficient capital markets: A review of theory and empirical work.Journal of Finance, 25(2), 383-417.
[4] C. I. Lee, K. C. Gleason & I. Mathur. (2001). Trading rule profits in Latin American currency spot rates. International Review of Financial Analysis, 10(2), 135-156.
[5] P. Saacke. (2002). Technical analysis and the effectiveness of central bank intervention. Journal of International Money and Finance, 21(4), 459-479.
[6] N. E. Huang, Z. Shen, S. R. Long, M. l. Wu, H. H. Shih, Q. Zheng, N. C. Yen, C. C. tung & H. H. Liu. (1998). The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis. Proceedings of the Royal Society of London.Series A:Mathematical, Physical and Engineering Sciences, 903-995.
[7] C. S. Lin, SH. Chiu & T. Y. Lin. (2012). Empirical mode decomposition-based least squares support vector regression for foreign exchange rate forecasting. Economic Modelling, 29(6), 2583-2590.
[8] J. A. K. Suykens & J. Vandewalle. (1999). Least squares support vector machine classifiers. Neural Processing Letters, 9(3), 293-300.
[9] W. Shu & Q. Gao. (2020). Forecasting stock price based on frequency components by EMD and neural networks. IEEE Access, 8, 206388-206395.
[10] G. E. P. Box & D. A. Pierce. (1970). Distribution of residual autocorrelations in autoregressive-integrated moving average time series models. Journal of the American Statistical Association, 65(332), 1509-1526.
[11] X. Tang, L. Wang, J. Cheng, J. Chen & V. S. Sheng. (2019). Forecasting model based on information-granulated GA-SVR and ARIMA for producer price index. Computers, Materials & Continua, 58(2), 463-491.
[12] A. Lapedes & R. Farber. (1987). Nonlinear signal processing using neural netsworks:prediction and system modeling. Technical Report LA-UR-87-2662, Los Alamos National Laboratory, Los Alamos, NM.
[13] W. Huang, K. K. Lai, Y. Nakamori & S. Wang. (2004). Forecasting foreign exchange rates with artificial neural networks: a review. International Journal of Information Technology and Decision Making, 3(1), 145-165.
[14] K.Shuang, R. Li, M. Gu, J. Loo & S. Su. (2020). Major-minor long short-term memory for word-level language model. IEEE Transactions on Neural Networks and Learning Systems, 31(10), 3932-3946.
[15] S. M. Alireza, E. Ali, P. Fernando & C. Paulo Lobato. (2021). Long short-term memory with gate and state level fusion for light field-based face recognition. IEEE Transactions on Information Forensics and Security, 16, 1365-1379.
[16] Q. Yu, K. Wang, J.O. Strandhagen & Y. Wang. (2018). Application of memory neural network to sales forecasting in retail—A case study. Advanced Manufacturing and Automation Vll, 451, 11-17.
[17] J. Qu, F. Liu, Y. Ma & J. Fan. (2019). A neural-network-based method for RUL prediction and SOH monitoring of lithium-ion battery. IEEE Access, 7, 87178-87191.
[18] B. Zhang. (2018). Foreign exchange rates forecasting with an EMD-LSTM neural networks model. Journal of Physics: Conf. Series,1053.
[19] J. Chung, C. Gulcehre, K.H. Cho & Y. Bengio. (2014). Empirical evaluation of gated recurrent neural networks on sequence modeling. In NIPS 2014 Workshop on Deep Learning.
[20] X. Ran, Z. Shan, Y. Fang & C. Lin. (2019). An LSTM-based method with
attention mechanism for travel time prediction. Sensors, 19(4), 861.
[21] H. Liu. (2018). Leveraging financial news for stock trend prediction with
attention-based recurrent neural network. arXiv:1811.06173, abs/1811.06173.
[22] J. Qiu, B. Wang, C. Zhou. (2020). Forecasting stock prices with long-short term memory neural network based on attention mechanism. PLOS ONE.
[23] R. Storn. (1996). On the usage of differential evolution for function optimization. In Proceedings of North American Fuzzy Information Processing, 519-523.
[24] J. Kennedy & R. Eberhart. (1995). Particle swarm optimization. Proceedings of ICNN'95 - International Conference on Neural Networks, 4, 1942-1948.
[25] D. Karaboga. (2005). An idea based on honey bee swarm for numerical optimization. Technical Report TR06, Erciyes University, Engineering Faculty, Computer Engineering Department.
[26] D. G. Mayer & B. P. Kinghorn & A. A. Archer. (2005). Differential evolution
an easy and efficient evolutionary algorithm for model optimisation.
Agricultural Systems, 83(3), 315-328.
[27] A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. u. Kaiser & I. Polosukhin. (2017). Attention is all you need. Paper presented at the 31th Conference on Neural Information Processing Systems, 5998-6008.
[28] X. Zhang, X. Liang, A. Zhiyuli, S. Zhang, R. Xu & B. Wu. (2019). AT-LSTM: An attention-based lstm model for financial time series prediction. Materials Science and Engineering, 569(5).
[29] R. J. Hyndman & G. Athanasopoulos. (2021). Forecasting: principles and practice. 3rd edition, OTexts: Melbourne, Australia. OTexts.com/fpp3.
[30] 陳旭昇、吳聰敏、李怡庭、陳虹宇。(2021)。致富的特權-二十年來我們為 央行政策付出的代價。春山出版，頁52-53。

網路文獻
金融時間序列的相位統計分析。網址: http://in.ncu.edu.tw/mcwu/publications/full_text/M_C_Wu_CPS_2MP_30_257(2008).pdf