全球化浪潮與科技日新月異推使計算機的計算效率提升，外加人工智慧、機器學習與深度學習等演算法崛起，使我們可以運用更先進的方法來解決問題，輔助決策制定。
本研究藉由利率、經濟數據、股市、匯率、金融情況等不同面向的數據，建立複迴歸模型(Multiple Regression Model)與長短期記憶神經網路模型(Long Short-Term Memory Networks Model)，欲預測實施低利率政策下美元計價的3個月LIBOR未來走勢。經實證結果顯示：第一，長短期記憶神經網路模型預測能力較複迴歸模型的預測能力好；第二，採用定錨式移動視窗法(Anchored Moving Window)時，若每一次預測的天數越少，則模型確度越高；第三，經隨機森林(Random Forest)挑選變數後的模型準確度低於或略低於全部變數，由此可驗證長短期記憶神經網路模型中解釋變數越多越好；第四，學習率並不是越高越好，將取決於目標變數，因此不同模型、資料有其合適的學習率。
本研究在實務層面上的貢獻不僅有利企業評價與投資報酬的決策，更能提升交易策略的勝率與金融衍生性商品的風險管理；在學術層面上的貢獻為本研究結合跨領域的知識，且目前極少論文探討神經網路運用於利率領域。因此，本研究欲探討長短期記憶神經網路預測利率的可行性與準確性。

Owing to globalization and the rapid progression of technology, the computational efficiency of computers has increased. The rising of algorithms, including artificial intelligence, machine learning and deep learning, enable us to utilize advanced methods to tackle problems and assist in decision-making.
In this study, I establish a multiple regression model and a long short-term memory neural network model to predict the future trend of 3-month LIBOR denominated in US dollars under a low interest rate policy by using data from different aspects, such as interest rates, economic data, stock market, exchange rates, financial situation, etc. The empirical results show that: first, the accuracy of the long short-term memory neural network model is better than that of the multiple regression model. Second, when the anchored moving window method is applied, the fewer days are predicted, the higher precision it will be. Third, compared to analyze with full variables, the accuracy is lower or slightly lower if the variables are selected by Random Forest. This result verifies that, in the long short-term memory neural network model, employing more explanatory variables is better. Last but not least, different models and materials have their own suitable learning rate.
This study aims at exploring the feasibility and the accuracy of long short-term memory neural networks in forecasting interest rates. In the practical aspect, this research benefit enterprises and stakeholders not only to facilitate business valuation and decision-making by expected return, but also to improve the winning rate of trading strategies and the risk management of derivatives. On the other hand, in the academic aspect, this master thesis serves as a pioneer to apply machine learning in the interest rate field via integrating neural networks into the knowledge of finance. Therefore, the contribution of this master thesis is significant.

Afonso, A., & Nunes, A. S. (2015). Economic forecasts and sovereign yields. Economic Modelling, 44, 319-326.

Akram, T., & Li, H. (2017). What keeps long-term US interest rates so low?. Economic Modelling, 60, 380-390.

Breuel, T. M. (2015). Benchmarking of LSTM networks. arXiv preprint arXiv:1508.02774.

Dewachter, H., & Lyrio, M. (2006). Macro factors and the term structure of interest rates. Journal of Money, Credit and Banking, 119-140

Duarte, J., Longstaff, F. A., & Yu, F. (2007). Risk and return in fixed-income arbitrage: Nickels in front of a steamroller?. The Review of Financial Studies, 20(3), 769-811.

François, C. (2017). Deep learning with Python.

Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural computation, 9(8), 1735-1780.

Kim, S. H., & Noh, H. J. (1997). Predictability of interest rates using data mining tools: a comparative analysis of Korea and the US. Expert Systems with Applications, 13(2), 85-95.

Kobor, A., Shi, L., & Zelenko, I. (2005). What determines US swap spreads?. The World Bank.

Lange, R. H. (2013). The Canadian macroeconomy and the yield curve: A dynamic latent factor approach. International Review of Economics & Finance, 27, 261-274.

Lekkos, I., Milas, C., & Panagiotidis, T. (2005). On the predictability of common risk factors in the US and UK interest rate swap markets: Evidence from non-linear and linear models. Department of Economics, Keele University.

Sarno, L., Thornton, D. L., & Valente, G. (2005). Federal funds rate prediction. Journal of Money, Credit and Banking, 449-471.

Swamynathan, M. (2019). Mastering machine learning with python in six steps: A practical implementation guide to predictive data analytics using python. Apress.

Tan, X. (2019, October). LIBOR Prediction Using Genetic Algorithm and Genetic Algorithm Integrated with Recurrent Neural Network. In 2019 Global Conference for Advancement in Technology (GCAT) (pp. 1-8). IEEE.

Vicente, J., & Tabak, B. M. (2008). Forecasting bond yields in the Brazilian fixed income market. International Journal of Forecasting, 24(3), 490-497.

Liang (2016). Python程式設計入門指南. 碁峯

Matthew Kirk (2017). 初探機器學習：使用Python. 歐萊禮

江庭瑀 (2018). 倫敦銀行同業拆借利率的問題現況與改革方向. 經貿法訓第226期，12-17

沈沛瑄、魏廉臻、張瑞益 (2019). 以LSTM-RNN預測ETF 50股價趨勢並結合交易策略以獲取最大獲利率.載於國立金門大學(主編). NCS 2019 全國計算機會議，36-41

袁麗梅 (2018). 應用機器學習預測台灣十年期公債殖利率. 國立台灣科技大學碩士學位論文

張力元 (2017). 深度學習應用於股價走勢之研究：以大陸市場為例.國立政治大學碩士論文

陳瑋光、賴明勇、林忠晶 (2009). 國外同業拆借利率期限結構影響因素的實證分析——以美元LIBOR為例[J]. 經濟數學，2009，26(02)：30-34.

華小嶽 (2017) . 中國境內美元同業拆借市場發展研究. 上海交通大學碩士學位論文

黃台心 (2009). 計量經濟學: Econometrics. 新陸書局

葉怡成 (2003). 類神經網路模式應用與實作. 儒林圖書

蔡立耑 (2018). 金融科技實戰：Python與量化投資. 博碩

鄧文淵 (2019). Python機器學習與深度學習特訓班：看得懂也會做的AI人工智慧實戰 = Machine learning and deep learning with python. 碁峯

簡劭騏 (2015). 主要國家央行採行負利率政策及啟示. 經濟研究第17期，273-306

簡禎富、許嘉裕 (2019). 大數據分析與資料挖礦. 前程文化