近年來，人工智慧技術發展迅速，類神經網路模型每年推陳出新，已被廣泛應用於時間序列資料之預測，其中證券市場更是受到投資人的關注，若能利用模型預測作為投資決策參考依據，結合自身的交易策略、未雨綢繆，便能有效對投資組合進行管理與規劃，以增加投資人獲利，也能夠避免不必要的損失。
本研究欲使用類神經網路模型架構對2004年2月至2022年5月的美國資訊科技類指數Vanguard Information Technology（VGT）的每日市場價格進行預測，將資料按80%、20%比例各分為訓練集與測試集，使用長短期記憶神經網路模型（LSTM）、時序卷積網路模型（TCN）、N-Beats模型、整合移動自我迴歸模型（ARIMA）進行預測，並以ARIMA模型做為基礎進行比較及分析。
最後，為得知各模型的預測能力，本文利用均方根誤差（RMSE）衡量預測結果。研究結果發現N-Beats模型相對於其他兩者類神經網路模型有更好的預測能力，且優於ARIMA模型。

In recent years, artificial intelligence technology has developed rapidly, and neural network models have been renewed every year, and have been widely used on predicting time series data. Such as the stock market, if investors can use the model to forecasting, reference for investment decisions and combine their own trading strategies, so that they not only can manage their investment portfolio effectively and increase investors’ profits but also avoid unnecessary losses. This study intends to use a neural network model to forecast the daily market price of the US information technology index Vanguard Information Technology (VGT) from February 2004 to May 2022. Divide the data into training set and test set according to the proportion of 80% and 20%, using Long Short-Term Memory neural network model (LSTM), Temporal Convolution Network model (TCN), N-Beats model, Autoregressive Integrated Moving Average model (ARIMA) for forecasting and comparison analysis, which based on the ARIMA model. Finally, in order to understand the predictive ability of each model, this study uses the Root Mean Square Error (RMSE) to measure the prediction results. The results of the study found that the N-Beats model not only has better predictive ability than the other two types of neural network models, but also better than the ARIMA model.

摘要-------------------------------------------------i
Abstract---------------------------------------------ii
目錄--------------------------------------------------iii
圖目錄------------------------------------------------v
表目錄------------------------------------------------vi
第一章 緒論-----------------------------------------------1
第一節 研究動機--------------------------------------------1
第二節 研究目的--------------------------------------------2
第二章 文獻回顧--------------------------------------------3
第一節 金融市場預測相關研究---------------------------------3
第二節 時間序列資料預測模型---------------------------------5
壹、整合移動平均自我迴歸模型（Autoregressive Integrated Moving Average model, ARIMA）----------------------------------------------5
貳、全連接神經網路（Fully-connect Neural Network, FNN）-----7
參、循環神經網路 (Recurrent neural network, RNN)------------8
肆、長短期記憶神經網路（Long Short-Term Memory, LSTM）-------9
伍、時序卷積網路（Temporal Convolution Network, TCN）-------11
陸、N-Beats模型--------------------------------------------14
第三章 研究設計及方法--------------------------------------17
第一節 研究架構-------------------------------------------17
第二節 研究方法-------------------------------------------20
壹、資料與分群---------------------------------------------20
貳、季節性檢定---------------------------------------------21
參、對數變換及差分處理--------------------------------------22
肆、模型選擇-----------------------------------------------23
伍、類神經網路資料正規化（Normalization）--------------------25
陸、類神經網路超參數優化（Hyper-parameters Optimization, HPO）25
第四章 研究結果-----------------------------------------------28
第一節 各模型預測結果視覺化----------------------------------28
第二節 比較各模型之均方根誤差（RMSE）--------------------------31
第三節 增加輸入資料之類神經網路模型效果------------------------32
第五章 結論與建議---------------------------------------------34
參考文獻-----------------------------------------------------36

Bai, S., Kolter, J. Z., & Koltun, V. (2018) An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling. arXiv:1803.01271
Box, G. and Jenkins, G. (1970) Time Series Analysis: Forecasting and Control. Holden-Day, San Francisco.
Burnham, K. P., & Anderson, D. R. (2004). Multimodel Inference: Understanding AIC and BIC in Model Selection. Sociological Methods & Research, 33(2), 261-304.
Campbell, J.Y., and Thompson, S. B. (2008). Predicting excess stock returns out of sample: Can anything beat the historical average? The Review of Financial Studies, 21, 1509-1531
Elman, J. L. (1990). Finding structure in time. Cognitive science, 14(2), 179-211.
Fama, E. F. (1965). The Behavior of Stock-Market Prices. The Journal of Business, 38(1), 34-105.
Fama, E. F. (1970). Efficient Capital Markets: A Review of Theory and Empirical Work, 25 J. FIN. 383.
Frank, M.Z., & Antweiler, W. (2001). Is All that Talk Just Noise? The Information Content of Internet Stock Message Boards. FEN: Behavioral Finance (Topic).
Glaria, B. A., et al. (1996) “Stock market indices in Santiago de Chile: forecasting using neural networks,＂IEEE International Conference on Neural Networks (4), 2172 -2175.
Granger, C. W. J., and Morgenstern, O. (1970). Predictability of stock market prices. Heath Lexington Books
Grudnistski, Gray and Osburn Larry, (1993), Forecasting S&P and Gold Futures Prices: An Application of Neural Networks, The Journal of Futures Markets 13(6), 631-643.
Harris, M., Raviv, A. (1993). Differences of Opinion Make a Horse Race. The Review of Financial Studies 6, 473-506.
Herzen, J., Lässig, F., Piazzetta, S.G., Neuer, T., Tafti, L., Raille, G., Pottelbergh, T.V., Pasieka, M., Skrodzki, A., Huguenin, N., Dumonal, M., Ko'scisz, J., Bader, D., Gusset, F., Benheddi, M., Williamson, C., Kosinski, M., Petrik, M., & Grosch, G. (2021). Darts: User-Friendly Modern Machine Learning for Time Series. ArXiv:2110.03224.
Hochreiter S, Schmidhuber J. (1997). Long short-term memory. Neural Comput. 15;9(8):1735-80.
Hyndman, R. J., & Khandakar, Y. (2008). Automatic Time Series Forecasting: The forecast Package for R. Journal of Statistical Software, 27(3), 1-22.
Jaffe, J., & Westerfield, R. (1985). The Week-End Effect in Common Stock Returns: The International Evidence. The Journal of Finance, 40(2), 433-454.
Jordan, M I. (1985-1986). Serial order: a parallel distributed processing approach. Technical report, United States.
Kenneth R. French. (1980). Stock returns and the weekend effect,Journal of Financial Economics,8(1),55-69
Keim DB, Stambaugh RF. (1984). A further investigation of the weekend effect in stock returns. The Journal of Finance, 39(3):819-835
Kingma, D. P. & Ba, J. (2014). Adam: a method for stochastic optimization. arXiv:1412.6980 .
Lakonishok, J., & Levi, M.D. (1982). Weekend Effects on Stock Returns: A Note. The Journal of Finance, 37, 883-889.
Liaw, R., Liang, E., Nishihara, R., Moritz, P., Gonzalez, J. E., Stoica, I.（2018）. Tune:A Research Platform for Distributed Model Selection and Training. ArXiv:1807.05118.
Lina Ni, Yujie Li, Xiao Wang, Jinquan Zhang, Jiguo Yu, Chengming Qi.(2019) Forecasting of Forex Time Series Data Based on Deep Learning, Procedia Computer Science, 147, 647-652
Marszałek, A., and Burczyński, T. (2014). Modeling and forecasting financial time series with ordered fuzzy candlesticks. Information sciences, 273 ,144-155.
Mills, T. C., and Markellos, R. N. (2008). The econometric modelling of financial time series. Cambridge University Press
Oreshkin, B.N., Carpov, D., Chapados, N., & Bengio, Y. (2020). N-BEATS: Neural basis expansion analysis for interpretable time series forecasting. ArXiv:1905.10437..
Pai,P.-F., & Lin,C.-S.(2005).A hybrid ARIMA and support vector machines model in stock price forecasting, Omega, 33, 497- 505.
Remy, P. (2020). https://github.com/philipperemy/n-beats
Rumelhart, D.E., Hinton, G.E., & Williams, R.J. (1986). Learning representations by back-propagating errors. Nature, 323, 533-536.
Wanjawa, B.W., & Muchemi, L. (2015). ANN Model to Predict Stock Prices at Stock Exchange Markets. ArXiv:1502.06434.
Yim, J. “A comparison of neural networks with time series models for forecasting returns on a stock market index,＂Lecture Notes in Computer Science 2002, 25-35
Zhao, Z., Rao, R., Tu, S., & Shi, J. (2017). Time-Weighted LSTM Model with Redefined Labeling for Stock Trend Prediction. 2017 IEEE 29th International Conference on Tools with Artificial Intelligence (ICTAI), 1210-1217.