採用二次回歸之趨勢值考慮多樣化的投資情形並結合具糾纏態之自適應量子啟發式禁忌搜尋演算法 解投資組合最佳化問題

Portfolio Optimization Considering Diversified Investment Methods Using ANGEL-QTS and Trend Ratio on Quadratic Regression

江育綺(Yu-Chi Jiang)

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

在股票市場裡，選擇投資組合可以有效地分散風險。好的評估策略能選擇具有良好表現的投資組合。夏普值是常見的評估指標之一，但它容易將走勢平穩的投資組合視為是最好的投資組合。本研究使用一次回歸的趨勢值來找出穩定上漲的投資組合，並且公平地比較不同區間長度下的投資組合。但一次回歸僅能粗略表達投資組合走勢，因此本研究提出二次回歸的趨勢值，二次曲線比一次線性更趨近投資組合的真實資金水位，也就能更加準確的評估投資組合的風險與報酬。另外，本研究提供多樣化的投資方法，像是定存、買賣張數或零股。定存能保本降風險並且獲得額外利息；張數及零股為台股市場常見的投資方法，在不同的投資情況下適合不同的投資方法及投資組合，因此本研究採用兩相的滑動視窗，除了能避免過度適應外，也能在多次的訓練中利用本研究提出的具糾纏態之自適應量子啟發式禁忌搜尋演算法(ANGEL-QTS)以更有效率地找出最適合的投資方法。由於投資組合常需多檔股票相互搭配才能表現更好，像是選了高報酬的股票就得同時搭配低風險的股票，而具糾纏態之自適應量子啟發式禁忌搜尋演算法能夠利用糾纏態的特性在最佳解附近進行區域搜尋，讓演算法嘗試多個位元的同時跳脫，也加強演算法的深度探勘能力。最後，實驗結果顯示本研究提出的方法比起夏普值、一次線性趨勢值，更能在訓練及測試區間中找到同時具有高報酬及低風險的良好投資組合。

關鍵字

選股 ； 投資組合 ； 趨勢值 ； 夏普值 ； 二次回歸 ； 趨勢線 ； 演算法 ； 區域搜尋 ； 零股 ； 定存 ； 滑動視窗

並列摘要

In the stock selection problem, the Sharpe ratio is one of the commonly used indicators, but it tends to identify the portfolio with a flat trend as the best one. This paper uses the linear trend ratio to access the portfolio with a stable upward trend. The trend ratio uses the trend line to assess the portfolio return and risk. However, the trend ratio with linear regression only roughly depict the portfolio trend. This paper proposes the quadratic trend ratio, utilizing the quadratic regression to find the portfolio trend curve. Therefore, the quadratic trend ratio can assess the portfolio return and risk more accurately. In addition to the access indicator, this paper provides diversified investments such as time deposit, buying round lots or buying odd lots. Different situation suits different investment method. Thus, this paper applies the 2-phase investment sliding windows to avoid the overfitting problem and chooses the best investment method by multiple training using Entanglement Local Search on Adaptive Global-best Guided Quantum-inspired Tabu Search with Not Gate (ANGEL-QTS) to find the best portfolio effectively and efficiently. Furthermore, the experimental results show that the proposed method can find the well-performing portfolio with higher return and lower risk in both the training and testing periods.

並列關鍵字

stock selection ； portfolio ； trend ratio ； Sharpe ratio ； quadratic regression ； trend line ； evolutionary algorithm ； local search ； odd lots ； time deposit ； sliding window

參考文獻

[1] W. F. Sharpe, “The sharpe ratio,” The Journal of Portfolio Management, vol. 21, no. 1, pp. 49–58, 1994.

[2] W. F. Sharpe, Investors and Markets: Portfolio Choices, Asset Prices, and Investment Advice. Princeton, NJ, USA: Princeton Univ. Press, 2011.

[3] H. Markowitz, “Portfolio selection,” The Journal of Finance, vol. 7, no. 1, pp. 77–91, 1952.

[4] Y. H. Chou, S. Y. Kuo and Y. C. Jiang, (in press). “A Novel Portfolio Optimization Model Based on Trend Ratio and Evolutionary Computation,” IEEE Transactions on Emerging Topics in Computational Intelligence, 2018.

[5] S. Y. Kuo, Y. C. Jiang, W. L. Yeoh and Y. H. Chou, “Portfolio Optimization Considering Diversified Investment Methods Using GNQTS and Trend Ratio,” IEEE International Conference on Systems, Man, and Cybernetics, 2018, pp. 3938–3943.