本研究旨在開發將機器學習和深度學習算法應用於投資回報預測和投資組合優化管理的系統，用於推薦投資者適當的短期和長期投資策略。我們設計的核心算法是一個人工智能嵌入式系統，用於在金融市場中進行時間序列預測，並專注於ETF交易。有許多研究側重於算法交易，傳統的時間序列預測和不同形式的各種應用的投資組合管理;然而，關注使用各種機器學習算法來預測市場趨勢的應用的文獻量是有限的。在這項研究中，我們使用了五種機器學習算法和兩種深度學習方法，長期短期記憶（LSTM）和閘循環單元（GRU）來開發我們的系統並構建短期和長期項目組合管理的應用程序。我們開發預測模塊並利用結果構建日間交易策略並執行投資組合優化以舉例說明機器學習算法真正增加價值，我們比較不同的方法並評估哪些算法在我們的任務中表現更好。我們構建的系統和模塊是可擴展和可移植的，它可以在不同的時間間隔內開發交易推薦系統時用作框架和子模塊。

There are many studies focus on algorithmic trading, traditional time series forecasting and portfolio management in different forms of various applications; however, the amount of literatures focusing on the applications which use various machine learning algorithm to forecast market trends are limited. In this research, we used five machine learning algorithms and two deep learning approaches, Long Short-Term Memory (LSTM) and Gate Recurrent Unit (GRU), to develop our system and build up the applications for short-term and long term portfolio management. We develop a forecasting module and exploit the result to construct a day trading strategy and to perform portfolio optimization to exemplify the machine learning algorithms truly add values, and we compare different methodologies and evaluate which algorithms perform better in our task. The system and module we built is expandable and portable, it can be used as a framework and submodule when developing trading recommendation system in different time intervals.

1.	Introduction	1
1.1.	Background and Motivation	1
1.2.	Research Purpose	3
1.3.	Research Method	3
1.4.	Conclusion and Findings	4
1.5.	Research Contribution	5
2.	Related Works	6
2.1.	Artificial Intelligence in Finance	6
2.2.	Machine Learning	8
2.2.1.	SVM	9
2.2.2.	Decision Tree	11
2.2.3.	eXtreme Gradient Boosting	12
2.2.4.	Random Forest	14
2.2.5.	Naïve Bayes	15
2.2.6.	Logistic Regression	16
2.3.	Neural Network and Deep Learning	17
2.3.1.	Recurrent Neuron Net (RNN)	19
2.3.2.	Long Short Term Memory (LSTM)	20
2.3.3.	Gate Recurrent Unit (GRU)	22
2.4.	Portfolio Optimization	24
2.4.1.	Markowitz Mean-Variance Model	25
2.4.2.	Black-Litterman model	25
2.4.3.	Summary of Related works on Portfolio Optimization	26
3.	Research Methods and System Framework	27
3.1.	Research Design	27
3.1.1	Construct a Conceptual Framework	27
3.1.2	Develop a System Architecture	27
3.1.3	Analyze & Design the System	28
3.1.4	Build the System	28
3.1.5	Observe & Evaluate the System	28
3.2.	System Architecture	29
3.2.1.	Data Collection Module	29
3.2.2.	Data Preprocessing Module	29
3.2.3.	Machine Learning Training Module	29
3.2.4.	Deep Learning Training Module	30
3.2.5.	Forecasting Module	30
3.2.6.	Data Visualization Module	30
3.2.7.	Portfolio Optimization Module	31
3.3.	Subjects and Data Collection	32
3.4.	Increase complexity of time series data and normalization	32
3.5.	Training and Validation	33
3.6.	Hyper parameter and Environment Setting	33
3.7.	The Black-Litterman Formula	34
4.	Experimental Result and Discussion	36
4.1.1.	Result and Analysis	37
4.1.2.	Comparison with Longer Time Horizon	40
4.1.3.	Portfolio Optimization With Model Based Investor Views	42
4.1.4.	Portfolio Assets Selection	43
5.	Conclusion and Recommendations	53
5.1.	Reviews and research finding	53
5.2.	Research contribution	54
5.3.	Practitioners implication	55
5.4.	Limitations of the Study	56
5.5.	Recommendations for Future Research	56
References	58




	
 
List of Figures
Figure 1. SVM classification	10
Figure 2. Classification Tree	11
Figure 3. Tree Ensemble Model	13
Figure 4. Random Forests	14
Figure 5. Logistic Regression compare to Linear Regression	16
Figure 6. Frank Rosenblatt’s Perceptron	18
Figure 7. A recurrent neural network and the unfolding in time.	19
Figure 8. A Long Short-Term Memory (LSTM) unit	21
Figure 9. The structure of GRU	22
Figure 10. System Development Research Method Flow Chart	28
Figure 11. System Architecture	31
Figure 12. Confusion Matrix for 0050 TW prediction in 2018	38
Figure 13. Volatility for 0050 TW prediction in 2018	40
Figure 14. Fluctuation for 0050 TW prediction in 2018	41
Figure 15. Correlation Matrix for whole 20 ETFs	43
Figure 16. Cumulative Return of 6 stocks (2018)	47
Figure 17. Cumulated return of portfolio optimization strategies(From Predict Model)	48
Figure 18. Cumulated return of portfolio optimization strategies (Return from Actual return)	49
Figure 19. Cumulated return (By Naïve Bayes) of portfolio optimization strategies	49
Figure 20. Cumulated return (By Logistic Regression) of portfolio optimization strategies	51
List of Tables
Table 1. Four approaches of AI	8
Table 2. A comparison of related works on stock market prediction	23
Table 3. A comparison of related works on portfolio optimization	26
Table 4. Structure of our deep learning model	37
Table 5. Predicted Result of 0050.TW by SVM	38
Table 6. Result and Performance of Each Predict Model	42
Table 7. Weight Rebalancing for Each Quarter	44
Table 8. List of Selected ETFs Date of Portfolio	44
Table 9. Weight of 6 ETFs of The Black-Litterman Model	45
Table 10. Correlation Matrix for the selected 6 ETFs	45
Table 11. Predicted Investors’ opinion matrix (Q1 and Q2 for instance)	46
Table 12. Correlation Matrix for the selected 6 ETFs	52

Black, F., & Litterman, R. (1990). Asset allocation: combining investor views with market equilibrium. Retrieved from 
Bourlard, H. A., & Morgan, N. (2012). Connectionist speech recognition: a hybrid approach (Vol. 247): Springer Science & Business Media.
Box, G. E., Jenkins, G. M., Reinsel, G. C., & Ljung, G. M. (2015). Time series analysis: forecasting and control: John Wiley & Sons.
Breiman, L. (2017). Classification and regression trees: Routledge.
Campbell, C., & Ying, Y. (2011). Learning with support vector machines. Synthesis lectures on artificial intelligence and machine learning, 5(1), 1-95. 
Chen, L., Qiao, Z., Wang, M., Wang, C., Du, R., & Stanley, H. E. J. I. A. (2018). Which Artificial Intelligence Algorithm Better Predicts the Chinese Stock Market? , 6, 48625-48633. 
Chen, T., & Guestrin, C. (2016). Xgboost: A scalable tree boosting system. Paper presented at the Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining.
Cho, K., Van Merriënboer, B., Bahdanau, D., & Bengio, Y. J. a. p. a. (2014). On the properties of neural machine translation: Encoder-decoder approaches. 
Chou, J.-S., & Nguyen, T.-K. (2018). Forward Forecast of Stock Price Using Sliding-Window Metaheuristic-Optimized Machine-Learning Regression. IEEE Transactions on Industrial Informatics, 14(7), 3132-3142. 
Chung, J., Gulcehre, C., Cho, K., & Bengio, Y. J. a. p. a. (2014). Empirical evaluation of gated recurrent neural networks on sequence modeling. 
Cortes, C., & Vapnik, V. J. M. l. (1995). Support-vector networks. 20(3), 273-297. 
Cox, D. R. (1958). The regression analysis of binary sequences. Journal of the Royal Statistical Society: Series B (Methodological), 20(2), 215-232. 
Engle, R. F. J. E. J. o. t. E. S. (1982). Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation. 987-1007. 
Glorot, X., Bordes, A., & Bengio, Y. (2011). Deep sparse rectifier neural networks. Paper presented at the Proceedings of the fourteenth international conference on artificial intelligence and statistics.
Hansson, M. (2017). On stock return prediction with LSTM networks. 
Hecht-Nielsen, R. (1992). Theory of the backpropagation neural network. In Neural networks for perception (pp. 65-93): Elsevier.
Ho, T. K. (1995). Random decision forests. Paper presented at the Proceedings of 3rd international conference on document analysis and recognition.
Hochreiter, S., & Schmidhuber, J. J. N. c. (1997). Long short-term memory. 9(8), 1735-1780. 
Idzorek, T. (2007). A step-by-step guide to the Black-Litterman model: Incorporating user-specified confidence levels. In Forecasting expected returns in the financial markets (pp. 17-38): Elsevier.
Jia-long, L., Bo-wei, L., & Min, L. (2013). Model contest and portfolio performance: Black-Litterman versus factor models. Paper presented at the 2013 International Conference on Management Science and Engineering 20th Annual Conference Proceedings.
Jiao, Y., & Jakubowicz, J. (2017). Predicting stock movement direction with machine learning: An extensive study on S&P 500 stocks. Paper presented at the 2017 IEEE International Conference on Big Data (Big Data).
Karolyi, G. A. J. J. o. B., & Statistics, E. (1995). A multivariate GARCH model of international transmissions of stock returns and volatility: The case of the United States and Canada. 13(1), 11-25. 
Kohavi, R. (1996). Scaling up the accuracy of naive-bayes classifiers: A decision-tree hybrid. Paper presented at the Kdd.
LeCun, Y. (2007). Who is afraid of non-convex loss functions. Paper presented at the NIPS Workshop on Efficient Machine Learning.
LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436-444. Retrieved from http://dx.doi.org/10.1038/nature14539. doi:10.1038/nature14539
Markowitz, H. J. J. o. p. E. (1952). The utility of wealth. 60(2), 151-158. 
Martellini, L., & Ziemann, V. (2007). Extending Black-Litterman analysis beyond the mean-variance framework. Journal of Portfolio Management, 33(4), 33. 
McCallum, A., & Nigam, K. (1998). A comparison of event models for naive bayes text classification. Paper presented at the AAAI-98 workshop on learning for text categorization.
McCarthy, J., Minsky, M. L., Rochester, N., & Shannon, C. E. J. A. m. (2006). A proposal for the dartmouth summer research project on artificial intelligence, august 31, 1955. 27(4), 12. 
McCulloch, W. S., & Pitts, W. J. T. b. o. m. b. (1943). A logical calculus of the ideas immanent in nervous activity. 5(4), 115-133. 
Ng, A. Y., Jordan, M. I., & Weiss, Y. (2002). On spectral clustering: Analysis and an algorithm. Paper presented at the Advances in neural information processing systems.
Nunamaker, J. F., Chen, M., & Purdin, T. D. M. (1990). Systems Development in Information Systems Research. Journal of Management Information Systems, 7(3), 89-106. doi:10.1080/07421222.1990.11517898
Paudel, R. B., & Koirala, S. (2006). Application of Markowitz and Sharpe Models in Nepalese Stock. Journal of Nepalese Business Studies, 3(1), 18-35. 
Quinlan, J. R. J. M. l. (1986). Induction of decision trees. 1(1), 81-106. 
Rather, A. M., Agarwal, A., & Sastry, V. J. E. S. w. A. (2015). Recurrent neural network and a hybrid model for prediction of stock returns. 42(6), 3234-3241. 
Rosenblatt, F. (1961). Principles of neurodynamics. perceptrons and the theory of brain mechanisms. Retrieved from 
Russell, S. J., & Norvig, P. (2016). Artificial intelligence: a modern approach: Malaysia; Pearson Education Limited.
Samuel, A. L. (1988). Some Studies in Machine Learning Using the Game of Checkers. II—Recent Progress. In Computer Games I (pp. 366-400): Springer.
Srivastava, T. (2015). Tuning the parameters of your Random Forest model. Analytics Vidhya, 9. 
Usmani, M., Adil, S. H., Raza, K., & Ali, S. S. A. (2016). Stock market prediction using machine learning techniques. Paper presented at the 2016 3rd International Conference on Computer and Information Sciences (ICCOINS).
van der Schans, M., & de Graaf, T. J. A. a. S. (2017). Robust optimization by constructing near-optimal portfolios. 
Vui, C. S., Soon, G. K., On, C. K., Alfred, R., & Anthony, P. (2013). A review of stock market prediction with Artificial neural network (ANN). Paper presented at the 2013 IEEE International Conference on Control System, Computing and Engineering.
Wang, H., Jiang, Y., & Wang, H. (2009). Stock return prediction based on Bagging-decision tree. Paper presented at the 2009 IEEE International Conference on Grey Systems and Intelligent Services (GSIS 2009).
Ye, X., Dong, C., Liu, T. J. S. S., & Systems. (2016). Image-based structural dynamic displacement measurement using different multi-object tracking algorithms. 17(6), 935-956. 
Yu, H., Chen, R., & Zhang, G. J. P. c. s. (2014). A SVM stock selection model within PCA. 31, 406-412. 
Zhao, L., & Palomar, D. P. (2018). A Markowitz Portfolio Approach to Options Trading. IEEE Transactions on Signal Processing, 66(16), 4223-4238.