運用多樣移動平均線及演化計算實現結合融券操作及停損機制的自動股票交易系統

Automatic Stock Trading System Combined with Short Selling and Stop Loss Using Multiple Moving Averages and EGQTS Algorithm

楊偉倫(Wei-Lun Yeoh)

指導教授 ： 周耀新

國立暨南國際大學/科技學院/資訊工程學系/碩士(2019年)

摘要

本研究提出了一種新型的自動股票交易系統，只使用一個簡單但常見的技術指標即移動平均線（Moving Average），並放寬參數以拓展其解空間。本方法也分析了加權移動平均（Weighted Moving Average）和指數移動平均（Exponential Moving Average）。此外，本方法使用經過改良的量子啟發式禁忌搜索演算法（Global best Quantum Tabu Search）以達到快速及穩定搜索合適的交易策略。且股票市場擁有大量歷史資訊，因太過久遠而失去參考價值，所以本方法結合滑動視窗在訓練-測試區間以不同長度進行實驗，並提出了隔年同期訓練-測試區間及分別持錢或持股的兩段式訓練，以解決更全面的股票交易問題。除了買低賣高的交易方式，交易系統還採用另一種合法的交易方式，融券。實驗結果表明，我們的方法改善了移動平均的能力；隔年同期訓練-測試及兩段式訓練的滑動視窗可以改善交易系統的表現；交易系統採用融券交易時可以顯著提高投資利潤。

關鍵字

自動化交易系統 ； 股票市場 ； 擇時 ； 技術分析 ； 移動平均線 ； 演化計算 ； 量子啟發式禁忌搜尋演算法 ； 滑動視窗

並列摘要

This research proposes a novel dynamic trading system utilizing only one simple but common technical indicator, the moving average (MA), albeit in a way which differs from traditional methods. We also analyze the weight moving average (WMA) and exponential moving average (EMA), which has the multiplying factor of MA. In addition, a modified evolutionary algorithm, the globe best-guide quantum-inspired tabu search algorithm (GQTS), was created to quickly and stably search for the optimal combination of MA parameters. In order to avoid the overfitting problem, this approach applied the sliding window, and raised the 2-phase sliding window and year-on-year training period to address more comprehensive stock trading problems. In addition to normal stock trading, this system adopts another legal trading method, short selling. The experiment results reveal that our method has a greatly improved MA ability. The result indicates that the WMA always has the best performance in four different targets. The sliding window period with 2-phase and year-on-year can improve the performance of the trading system. When the trading system adopts short selling it can significantly improve investment profit.

並列關鍵字

Automated trading system ； Stock market ； Market timing ； Technical analysis ； Moving average ； Evolutionary algorithm ； Quantum inspired tabu search algorithm ； Sliding window

參考文獻

[1] N. Kourentzes, and F. Crone, “Frequency independent automatic input variable selection for Neural Networks for forecasting”, Neural Networks (IJCNN), The 2010 International Joint Conference on, pp.1-8, 2010.

[2] C. James, "Kanban Cell Neuron network Stock Trading System (KCNSTS)," 2015 International Joint Conference on Neural Networks (IJCNN), Killarney, 2015, pp. 1-8.

[3] C. -C. Chen, C. Kuo, S. -Y. Kuo, and Y. -H. Chou, “Dynamic Normalization BPN for Stock Price Forecasting”, Systems, Man, and Cybernetics (SMC), 2015 IEEE International Conference on, pp. 2855-2860, 2015.

[4] H. Huang, M. Pasquier, and C. Quek, “Financial market trading system with a hierarchical coevolutionary fuzzy predictive model”, IEEE transactions on Evolutionary Computation, vol. 13, no. 1, pp. 56-70, 2009.

[5] I. -C. Yeh, and C. -H. Lien, ‘‘Fuzzy rule-based stock trading system”, Fuzzy Systems (FUZZ), 2011 IEEE International Conference on, pp. 2066–2072, 2011.