股市預測於學術界收到大量的關注，用於預估公司之未來價值已用於金融交易的決策。本研究之目標為結合深度學習與統計方法以用於預測股價。但股市預測大多牽涉非線性、多變及複雜的資料。本研究之複合模型結合離散小波變換(Discrete Wavelength Transform (DWT))、疊式稀疏自動編碼器(Stacked Sparse Autoencoder (SSAE))與隱馬可夫模型(Hidden Markov Model (HMM))。DWT用於清理資料雜訊，以SSAE選擇特徵，並以HMM預測股市指數走向。本研究以八年TAIEX資料為實驗數據以測試此複合式模型，並以(Mean Absolute Percentage Error (MAPE))、方均根差(Root Mean Square Error (RMSE))、R與Theil U為判斷之準則。近年演算法交易演進迅速，本研究提供此新複合預測模型並以實際資料驗證其預測效力。本研究並以Nifty 50 stock index之資料與過去研究劑型比較，本研究之複合模型在各時段皆有較好的表現，顯示結合深度學習與統計複合式模式與統計方法用於預測金融時間序列的效率。

Stock market prediction is the act of trying to determine the future value of a company stock or other financial instrument traded on an exchange. This act has attracted much attention from academia. However, as a results of its non-linear, volatile and complex nature of the data, it is quite difficult to predict. Thus the motivation behind this research is to innovatively combine a novel deep learning and statistical framework where Discrete Wavelength Transform (DWT), Stacked Sparse Autoencoder (SSAE) and Hidden Markov Model (HMM) are combined for stocked price forecasting. Three refined processes have been propose in the hybrid model for forecasting: 1) Use DWT to eliminate noise from the data via decomposition; 2) Select highly relevant features via SSAE; 3) Employ a Hidden Markov Model to predict the stock market price trend. An eight year period of Taiwan Stock Exchange Capitalization Weighted Stock Index (TAIEX) historical data is employed as experimental database to test the prediction ability of the proposed model with a performance indicator; Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), R and Theil U. Algorithm trading has evolved exponentially in recent years, thus, this research offers evidence on the predictive ability of a new hybrid forecasting model. To evaluate the robustness of model, Nifty 50 stock index data was also used and when compared to other state-of-the-art models, it outperforms them based on the same time frame and market condition. This research has demonstrated that combination of deep learning and statistical method are effective in the prediction of financial time series data.

摘要 iii
ABSTRACT iv
Acknowlegdement v
Table of Contents vii
List of Tables ix
List of figures x
Chapter 1 - Introduction 1
1.1 Background of Study 1
1.2 Objectives 4
1.3 Problem description and research significance 5
1.4 Research Hypothesis 6
Chapter 2 – Literature Review 8
2.1 Fundamental Analysis 8
2.2 Technical Analysis 8
2.3 Machine, Deep learning, & Hidden Markov models in Stock market prediction 9
Chapter 3 – Research Methodology 14
3.1 Model Specification 16
3.1.1 Data preprocessing using Discrete Wavelet Transform 16
3.1.2 Feature extraction using Stacked Sparse Autoencoder 18
3.1.3 Hidden Markov Model for prediction 20
3.2 Data Description 22
3.3 Prediction Approach 22
3.1.3 Performance Measurement 24
Chapter 4 – Results and Analysis 26
Chapter 5 – Main Findings 36
5.1 Conclusion and Future work 37
Bibliography 39

 

[1] Reference One et al., Phys. Rev. C0, 999 (2000).
[2] A. Popoola, K. A. (2006). Testing the Suitability of Wavelet Preprocessing for TSK
Fuzzy Models. IEEE: International Conference Fuzzy System Networks, 1305-1309. Doi: 10.1109/FUZZY.2006.1681878
[3] Akhter Mohiuddin Rathera, A., V.N.Sastry. (2015). Recurrent neural network and a hybrid model for prediction of stock returns. Expert Systems with Applications, 42(6), 3234-3241.
[4] Anastasis Brovykh, S. B., and Cornelis W. Oosterlee. (2017). Conditional Time Series Forecasting with Convolutional Neural Networks.
[5] Charles D. Kirkpatrick, J. R. D. (2006). Technical Analysis: The Complete Resource for Financial Market Technicians. Financial Time Press.
[6] Chat_eld, C. (2006). What is 'best' method for forecasting, Journal of Applied Statistics, 15(1). doi: http://dx.doi.org/10.1080/02664768800000003
[7] Dalto, M. (2017). Deep neural networks for time series prediction with applications in ultra-short-term wind forecasting.
[8] Honchar, L. D. P. a. O. (2016). Arti_cial Neural Networks architectures for stock price prediction: comparisons and applications. International Journal of Circuits, System and Signal Processing, 10.
[9] Hui-KuangYu. (2005). Weighted fuzzy time series models for TAIEX forecasting. PhysicaA: Statistical Mechanics and its Applications, 349(3-4), 609-624.
[10] J. G. Agrawal, D. V. S. C., Dr. A. K. Mittra. (2013). State-of-the-Art in Stock Prediction Techniques. International Journal of Advanced Research in Electrical, Electronics and Instrumentation Engineering.
[11] Jiawei Han , M. K. (2006). Data Mining: Concepts and Techniques (2nd Edition). Morgan Kaufmann, San Francisco.
[12] Kai Chen, Y. Z., Fangyan Dai. (2015). A LSTM-based method for stock returns prediction: A case study on China stock market. IEEE International Conference on Big Data.
[13] Karaboga, D. O., Celal. (2009). Neural networks training by arti_cial bee colony algorithm on pattern classi_cation. Neural Network World, 19, 279-292.
[14] Liang-Ying Wei, T.-L. C. a. T.-H. H. (2011). A hybrid model based on adaptive-network- based fuzzy infeence system to forecast taiwan stock market. Expert Systems with Applications, 38, 13625-13631.
[15] Ma, Y. L. a. W. (2010). Applications of Arti_cial Neural Networks in Financial Economics: A Survey. IEEE Computer Society Washington, 01, 211-214. doi: 10.1109/ISCID.2010.70
[16] Mingyue Qiu, Y. S., Fumio Akagi. (2016). Application of arti_cial neural network for the prediction of stock market returns: The case of the Japanese stock market. Chaos, Solitons and Fractals, 85, 1-7.
[17] Nijole Maknickiene, A. V. R., Algirdas Maknickas. (2011). Investigation of financial market prediction by recurrent neural network. Innovative Infotechnologies for science, Business and Education, 2(11), 3-8.
[18] Pimwadee Chaovalit, A. G., George Karabatis, Zhiyuan Chen. (2011). Discrete wavelet transform-based time series analysis and mining. ACM Computing Surveys (CSUR), 43(2).doi: 10.1145/1883612.1883613
[19] Qing Li, Y. C., Li Ling Jiang, Ping Li and Hsinchun Chen. (2016). A Tensor-Based Information Framework for Predicting the Stock Market. ACM Transactions on Information Systems 34 (2). doi: http://dx.doi.org/10.1145/2838731
[20] Ramsey, J. B. (1999). The contribution of wavelets to the analysis of economic and financial data. Philosophical Transactions of the Royal Society A- Mathematical, Physical and Engineering Sciences, 357(1760). doi: 10.1098/rsta.1999.0450
[21] Rodolfo C. Cavalcante, R. C. B., Victor L.F Souza, Jaeley P. Nobrega and Adriano L.I. Oliveira. (2016). Computational Intelligence and Financial Markets: A survey and Furure Directions. Expert Systems with Applications, 55, 194-211. doi: http://dx.doi.org/10.1016/j.eswa.2016.02.006
[22] S. Kumar Chandar, M. S. a. S. N. S. (2016). Prediction of stock market price using hybrid of wavelet transfrom and arti_cial neural network. Indian Journal of Science and Technology, 9(8). doi: 10.17485/ijst/2016/v9i8/87905
[23] Sepp Hochreiter, J. S. (1997). Long Short-Term Memory. Neural Computation, 9(8), 1735-1780.
[24] Shyi-MingChen. (1996). Forecasting enrollments based on fuzzy time series. Fuzzy Sets Systems, 81(3), 311-319. doi: https://doi.org/10.1016/0165-0114(95)00220-0
[25] Specht, D. F. (1990). Probabilistic neural networks. Neural Networks, 3(1), 109 - 118.
[26] Subasi, A. (2005). Epileptic Seizure detection using Dynamic Wavelet Network. Expert Systems with Applications, 29, 343-355.
[27] Tsung-Jung Hsieh, H.-F. H. a. W.-C. Y. (2011). Forecasting stock markets using wavelet transforms and recurrent neural networks: An integrated system based on artificial bee colony algorithm. Applied Soft Computing, 11, 2510-2525. doi: 10.1016/j.asoc.2010.09.007
[28] Xiao Ding, Y. Z., Ting Liu and Junwen Duan. (2015). Deep Learning for Event-Driven Stock Prediction. International Joint Conference on Artificial Intelligence (IJCAI).
[29] Xin-Yao Qian, S. G. (2017). Financial Series Prediction: Comparison between Precision of Time Series Models and Machine Learning Methods.
[30] Yingjun Chen, Y. H. (2017). A feature weighted support vector machine and K-nearest neighbor algorithm for stock market indices prediction. Expert Systems with Applications,80, 340-355.
[31] Yoshua Bengio, P. L., Dan Popovici, Hugo Larochelle. (2006). Greedy layer-wise training of deep networks. ACM Computing Surveys (CSUR), 1.
[32] Zachary C. Lipton, J. B., Charles Elkan. (2015). A Critical Review of Recurrent Neural Networks for Sequence Learning. arXiv preprint arXiv. doi: arXiv:1506.00019
[33] Li Z, Wu Z, He Y and Fulei (2005), Hidden Markov Model-based fault diagnostic method in speed-up and speed-down process for rotating machinery. Mechanical Systems and Signal Processing, vol. (19) 2, pp. 329-339.
[34] Xie. H, Anreae P, Zhang M, Warren P (2004), learning models for English Speech Recognition, Proceedings of the 27th Conference on Australasian Computer Science, pp. 323-329.
[35] Liebert M A (2004), Use of runs statistics for pattern recognition in genomic DNA sequences. Journal of Computational Biology, Vol. 11, pp. 107-124.
[36] Vinciarelli A and Luettin J (2000), Off-line cursive script recognition based on continuous density HMM, Proceedings of the 7th International Workshop on Frontiers in Handwriting Recognition, Amsterdam, pp. 493-498.
[37] M.R Hassan (2009). A combination of hidden markov model and fuzzy model for stock market forecasting. Journal of Neurocomputing, pp. 3439-3446.
[38] A.S Weigend and S. Shi (2000). Predicting daily probability distributions of S&P500 returns, Journal of Forecasting, pages 375-392.
[39] A. Gupta, B. Dhingra, (2012). Stock market prediction using Hidden Markov models, IEEE.
[40] Md. R. Hassan, B. Nath, M, Kirley, (2006). A fusion model of HMM, ANN & GA for stock market forecasting, Expert System with Applications, 33, pp. 171-180, doi:10.1016/j.eswa.2006.04.007.
[41] W. Bao W, Yue J, Rao Y (2017) A deep learning framework for financial time series using stacked autoencoders and long-short term memory. PLoS ONE 12(7): e0180944. https://doi. org/10.1371/journal.pone.0180944.
[42] Guo Z, Wang H, Liu Q, Yang J. A feature fusion based forecasting model for financial time series. Plos One. 2014; 9(6), 172-200.
[43] Hsieh TJ, Hsiao HF, Yeh WC. Forecasting stock markets using wavelet transforms and recurrent neural networks: An integrated system based on artificial bee colony algorithm. Applied soft computing. 2011; 11(2): 2510-25.
[44] EK. O.. Forteacsting Nigerian Stock Exchange Retursn: Evidence from Autoregressive integrated Moving Average (ARIMA) model. Ssrn Electronic Journal. 2010.
[45] Bliemel F. Theil’s Forecast. Accuracy Coefficient: A clarification. Journal of Marketing Research. 1973: 10 (4): 444.
[46] Chen, S. M. (1996). Forecasting enrollments based on fuzzy time-series. Fuzzy SetsSystems, 81, 311–319.
[47] Yu, H. K. (2005). Weighted fuzzy time-series models for TAIEX forecasting. PhysicaA, 349, 609–624.
[48] Wei L. Y, Chen T.L, Ho. T. H. (2011). A hybrid model based on adaptive-network-based fuzzy inference system to forecast Taiwan stock market. Expert Systems with Applications, 38 (2011) 13625–13631.