臺灣博碩士論文加值系統

世界經濟論壇在2015年指出金融科技(Fintech)將會是一個破壞性的創新，會全面改造金融產業的未來。其中，在智慧理財領域，Business Insider提出全世界被理財機器人所管理的資金將從2015年的一千億美元上漲到2020的八兆美元。相較於電腦，透過人力做投資會存在三大劣勢，包含記憶能力、決策速度，及情緒管理能力。因此本篇論文將建立一個理財機器人藉此提升交易的競爭能力。
本篇論文的理財機器人不同之處在於未平倉量的處理方式。過去文獻都將未平倉的數量直接作為輸入變數，本篇論文則是將未平倉量轉換成金額單位建立能夠同時評估現貨、期貨、與選擇權未平倉量的綜合指標作為輸入變數。在做投資決策時，若沒有同時評估外資衍生性金融商品的未平倉量，很容易被外資現貨的進出狀況誤導，做出錯誤決策。在訓練模型上，我們將使用Q-learning演算法作為訓練模型，並透過所建立綜合指標以及技術分析資料建立State，以台灣期貨指數的賺賠金額建立Reward。並利用台灣加權指數作為訓練市場，實作兩個SSCI期刊中的金融方法，分別為MACD以及EMA方法。最後再以Sharpe ratio來當作三個方法的評估依據，得到利用Q-learning演算法去做交易策略能獲得較高的Sharpe ratio.

The World Economic Forum pointed out in 2015 that Fintech would be a disruptive innovation that will revolutionize the future of the financial industry. In the field of investment management, Business Insider proposed that the world's funds managed by Robo-Advisors will rise from USD 100 billion in 2015 to USD 8 trillion in 2020. Investing through humans have three disadvantages compared to computers, including memory, the speed of making decision, and the ability of emotion management. Therefore, this study established a Robo-Advisor to enhance the competitiveness of investing.
The novelty of the proposed Robo-Advisor is that we convert the volume of open interest into a unit of value as an input variable. When making investment decisions, if not assessing the open interest of foreign investment institution in different financial derivatives at the same time, it is easy for investor to confuse by foreign investment institution to make the wrong decision. Therefore, we will establish a comprehensive evaluation indicator of open interest as the input variable of Q-learning algorithm. We define the states by the comprehensive indicator and technical analysis data, and define the rewards by the profit of trading on(The future of Capitalization Weighted Stock Index) TX. We use the Taiwan Capitalization Weighted Stock Index as a training market, and implement two financial methods including moving average convergence divergence (MACD), and exponential moving average (EMA). Finally, we used the Sharpe ratio as the basis for the evaluation of the three methods, then the Q-learning algorithm achieve a higher Sharpe ratio.

CONTENTS
摘要 I
Abstract II
誌謝 III
CONTENTS V
List of Tables VII
List of Figures VIII
Chapter 1 Introduction 1
1.1 Research Background 1
1.2 Research Motivation 3
1.3 Research Objectives 4
1.4 Research Process 5
Chapter 2 Literature Review 7
2.1 Candlestick Chart 7
2.2 Open Interest 8
2.3 The Pricing Model of Financial Derivatives 10
2.3.1 Cost of Carry Model 10
2.3.2 The Options Pricing Model of Black and Scholes 11
2.4 The Strategy of Trading 12
2.4.1 Bollinger Bands 12
2.4.2 Way of the Turtle 13
2.5 Model of Forecasting Stock Market 14
2.5.1 Related Studies and Comparison 14
2.5.2 Reinforcement Learning 15
2.5.3 Q-learning Algorithm 16
2.6 Summary 18
Chapter 3 Research Methodology 19
3.1 Problem Definition 19
3.2 Research Method Flow 19
3.3 Data Preprocessing 21
3.4 Construct the Comprehensive Indicator of Open Interest 21
3.4.1 HyperParameter 22
3.4.2 Evaluate the Profit of TAIEX 23
3.4.3 Evaluate the Profit of Future 24
3.4.4 Evaluate the Profit of Option 24
3.4.5 Comprehensive Indicator of Open Interest 25
3.5 Robo-Advisor Design 26
3.5.1 State Space 26
3.5.2 State Definition 26
3.5.3 Action Definition 29
3.5.1 Reward Definition 29
3.5.2 The Algorithm of Robo-Advisor 30
Chapter 4 Experiment and Analysis 31
4.1 Experimental Architecture 31
4.2 Experiment Implementation 32
4.2.1 Data Set Description 32
4.2.2 Hyperparameter Setting 34
4.3 Baseline methods 34
4.3.1 Exponential Moving Average 34
4.3.2 Moving Average Convergence Divergence 35
4.4 Experimental Results and Analysis 35
4.4.1 Q-table Analysis 36
4.4.2 Trading Performance Analysis 38
Chapter 5 Conclusion and Future work 41
5.1 Conclusion 41
5.2 Managerial Implication 42
5.3 Future work and Limitation 43
5.3.1 Limitation 43
5.3.2 Future work 43
References 44

Adebiyi, A. A., Adewumi, A. O., ＆ Ayo, C. K. (2014). Comparison of ARIMA and Artificial Neural Networks Models for Stock Price Prediction. Journal of Applied Mathematics, 2014, 1-7.

Almahdi, S., ＆ Yang, S. Y. (2017). An adaptive portfolio trading system: A risk-return portfolio optimization using recurrent reinforcement learning with expected maximum drawdown. Expert Systems with Applications, 87, 267-279.

Bates, R. G., Dempster, M. A., ＆ Romahi, Y. S. (2003). Evolutionary reinforcement learning in FX order book and order flow analysis. Proceedings of the Computational Intelligence for Financial Engineering, 355-362.

Bhuyan, R., ＆ Chaudhury, M. (2005). Trading on the information content of open interest: Evidence from the US equity options market. Derivatives Use, Trading ＆ Regulation, 11(1), 16-36.

Black, F., ＆ Scholes, M. (1973). The pricing of options and corporate liabilities. The Journal of Political Economy, 81(3), 637-654.

Chen, W. H., Shih, J. Y., ＆ Wu, S. (2006). Comparison of support-vector machines and back propagation neural networks in forecasting the six major Asian stock markets. Electronic Finance, 1(1), 49-67.

Cho, S., Cho, S., ＆ Yow, K. C. (2017). A robust time series prediction model using POMDP and data analysis. Advances in Information Technology, 8, 154-158.

Cornell, B., ＆ French, K. R. (1983). The pricing of stock index futures. Futures markets, 3(1), 1-14.

Day, M. Y., Cheng, T. K., ＆ Li, J. G. (2018). AI Robo-Advisor with Big Data Analytics for Financial Services. Proceedings of the Advances in Social Networks Analysis and Mining (ASONAM), 1027-1031.

Gao, X., ＆ Chan, L. (2000). An algorithm for trading and portfolio management using Q-learning and sharpe ratio maximization. Proceedings of the Neural information processing, 832-837.

Houlihan, P., ＆ Creamer, G. G. (2017). Can Sentiment Analysis and Options Volume Anticipate Future Returns? Computational Economics, 50(4), 669-685.

Hsu, H., ＆ Wang, J. (2004). Price expectation and the pricing of stock index futures. Review of Quantitative Finance and Accounting, 23(2), 167-184.

Huang, S. C., ＆ Wu, T. K. (2008). Integrating GA-based time-scale feature extractions with SVMs for stock index forecasting. Expert Systems with Applications, 35(4), 2080-2088.

Kamble, R. A. (2017). Short and long term stock trend prediction using decision tree. Proceedings of the Intelligent Computing and Control Systems (ICICCS), 1371-1375.

Kannan, K. S., Sekar, P. S., Sathik, M. M., ＆ Arumugam, P. (2010). Financial stock market forecast using data mining techniques. Proceedings of the International Multiconference of Engineers and computer scientists.

Kocianski, S. (2016). THE ROBO-ADVISING REPORT: Market forecasts, key growth drivers, and how automated asset management will change the advisory industry. Retrieved November 7, 2018 from https://nordic.businessinsider.com/the-robo-advising-report-market-forecasts-key-growth-drivers-and-how-automated-asset-management-will-change-the-advisory-industry-2016-6/.

Lai, H.-C., Tseng, T.-C., ＆ Huang, S.-C. (2016). Combining value averaging and Bollinger Band for an ETF trading strategy. Applied Economics, 48(37), 3550-3557.

Lee, J. W., Park, J., Jangmin, O., Lee, J., ＆ Hong, E. (2007). A Multiagent Approach to Q-Learning for Daily Stock Trading. IEEE Transactions on Systems, Man, and Cybernetics - Part A: Systems and Humans, 37(6), 864-877.

Lee, K., ＆ Jo, G. (1999). Expert system for predicting stock market timing using a candlestick chart. Expert Systems with Applications, 16(4), 357-364.

Lehoczky, J., ＆ Schervish, M. (2018). Overview and History of Statistics for Equity Markets. Annual Review of Statistics and Its Application, 5, 265-288.

Li, B., Zhang, D., ＆ Zhou, Y. (2017). Do trend following strategies work in Chinese futures markets? The journal of futures markets, 37(12), 1226-1254.

Moody, J., Wu, L., Liao, Y., ＆ Saffell, M. (1998). Performance functions and reinforcement learning for trading systems and portfolios. Forecasting, 17(5‐6), 441-470.

Neuneier, R. (1998). Enhancing Q-learning for optimal asset allocation. Proceedings of the Advances in neural information processing systems, 936-942.

Pendharkar, P. C., ＆ Cusatis, P. (2018). Trading financial indices with reinforcement learning agents. Expert Systems with Applications, 103, 1-13.

Seyhun, H. N. (1986). Insiders' profits, costs of trading, and market efficiency. Financial Economics, 16(2), 189-212.

Yong, L., Bo, Z., ＆ Yong, T. (2015). Dynamic optimal capital growth of diversified investment. Journal of Applied Statistics, 42(3), 577-588.