考量投資標的基本面、技術面分析並運用遺傳基因演算法決定最佳投資組合

Using Genetic Algorithm to Determine the Optimal Portfolio with Fundamental and Technical Analysis

謝宏鑫(Hung-Hsin Hsieh)

指導教授 ： 吳泰熙

國立臺北大學/商學院/企業管理學系/碩士(2009年)

摘要

1952年Markowitz提出投資組合的概念，奠定了投資研究領域的基礎。爾後隨著投資標的增加、數學模型及交易機制的變化，使得投資研究不再只是單純的數學問題；衍生而成的是一種結合傳統投資理論、科學模型與投資市場相關訊息的整合性研究。
近代投資研究領域中，運用各類型的人工智慧演算法，例如：遺傳基因演算法(Genetic Algorithm，簡稱GA)，在決定最佳投資組合的議題上有眾多的貢獻。於許多以GA求解投資組合的研究當中，屬指數追蹤的研究最為廣泛。然而，上述的研究卻忽略了投資人除了以過去績效和風險，選擇投資標的外，仍重視投資標的之基本面資訊。另外，技術分析亦是研究時必須考慮的要因之一。故本研究除了運用GA決定最佳投資組合外，亦考量投資標的之基本、技術面分析，試圖模擬投資決策的步驟與考量的資訊。
　　本研究建構的模型主要概念來自於夏普指標，但由於夏普指標考量的風險為變異數，故本研究將採用與實務投資關聯性較高的半變異數進行修正。實證階段中，本研究以所提出之修正夏普指標建構五大投資模式，於2005至2008年為投資期間進行實證，並以台灣50指數之績效為標竿進行比較。結果顯示考量基本分析，並以修正夏普指標最大化為決策目標，運用GA決定權重之投資組合，在投資期間可獲得平均年報酬率17.11%的績效，而台灣50指數於同一期間僅獲得-3.1%的平均年報酬。顯示本研究建構之投資模式於投資期間內，獲利能力優於台灣50指數之表現。

關鍵字

基因演算法 ； 投資組合 ； 基本分析 ； 技術分析 ； 下方標準差

並列摘要

In 1952, Markowitz innovates the mean-variance model for the portfolio selection problems. Henceforward this model has been the cornerstone of portfolio research and has served as a basis for the development of financial investment methodology. However, the problems of modern investment have become more complex, so the conventional methodology seems incapable of solving those problems effectively. Therefore, it is a very important issue that how to acquire the problem-solving methods between the computational technology and the traditional model.
During the last decade, using artificial intelligence to solve portfolio problems has become a creative trend in investment research and one of those popular artificial intelligence is the genetic algorithm (GA). In many portfolio researches using GA, the index tracking problem is the most common subject. However, the fundamental and technical analysis, which have been emphasized criteria in stock investment, have been ignored. Meanwhile, because the Sharpe Index can not reflect the investment loss literally, so we revise the Sharpe Index through the Semi-Deviation.
Therefore, this paper presents a modified Sharpe Index model and applies genetic algorithm on the optimal portfolio selection problems, in which the investments’ fundamental and technical analysis are considered and performed. Meanwhile, we propose five investment procedures, and those proposed investment procedures have been tested with TSEC Taiwan 50 Index’s data from 2005 to 2008. The empirical results show the one of proposed investment procedures (Procedure 3) is able to obtain approximately 17.11% average rate of return. Furthermore, the results also dominate the performance of TSEC Taiwan 50 Index during the same empirical period. Besides, the synergy of combining the rule of experience and scientific models has been observed in this paper.

並列關鍵字

Genetic algorithm ； Portfolio ； Fundamental analysis ； Technical analysis

參考文獻

吳詩敏（2007），組合編碼遺傳演算法於投資策略資金分配之應用，國立中央大

施正宏（2004），結合總體經濟指標及個股財務資料以預測個股漲跌－以台灣電

黃光廷（2002），技術分析、基本分析與投資組合避險績效之研究，國立成功大

Allen, F., Karjalainen, R. (1999). Using genetic algorithms to find technical trading rules. Journal of Finance Economics, 51, 245-271.

Beaver, W. H. (1968). The information content of annual earnings announcements. Journal of Accounting Research, 6, 67-92.