透過您的圖書館登入
IP:3.143.168.172
  • 學位論文

結合決策樹與遺傳演算法建構不同風險程度之基金投資組合-以國內發行之股票基金為例

Combine Decision Tree and Genetic Algorithms to Construct Mutual Fund Portfolio Based on Perceived Risk Levels

指導教授 : 皮世明

摘要


共同基金為國內投資人經常使用的投資理財工具,但大多數的投資人侷限於專業能力之不足,往往無法選擇到符合期待的基金。故本研究將結合決策樹與遺傳演算法來建置不同風險程度的基金投資組合,利用決策樹的分類技術找出影響投資組合風險的顯著因素,並透過遺傳演算法的搜尋與最佳化能力,依據投資者需求建構不同風險程度之投資組合。 本研究以β Coefficient做為基金風險程度分類的依據,而在選擇基金時,以基金的歷史報酬率為主,另外以Standard Deviation、β Coefficient、Sharpe Index、Jensen Index、Treynor Index和Information Ratio等六個整體績效指標,與七個會影響基金績效之基金特性,包括週轉率,費用率,基金規模,經理人是否擁有MBA學位,投信規模、基金是否獲得過獎項與基金經理人是否獲獎,當作選擇投資組合內基金的評估準則,並以二元編碼法作為遺傳演算法的編碼方式。資金配置的部分,乃利用遺傳演算法所產生的基金投資組合中,個別基金所佔的比例為之。最後,二元編碼染色體,利用各績效指標並搭配決策樹歸納出影響風險指標所建構出之適應值函數來評估,再經由遺傳演算法之演化過程,建立出高報酬或低風險的基金投資組合。 經過實驗後證實將決策樹所歸納出影響風險的因素納入遺傳演算法,可以獲取較好的績效。其中風險程度最高的R3型基金頭資組合,有最顯著的影響。

並列摘要


Mutual fund is one of the more popular financial products of Taiwanese investors; but due to the lack of professional expertise in investing, very seldom can one choose a fund the meets the expectation of the investor. The aim of this research is to combine decision tree and Genetic algorithms to construct mutual fund portfolio based on perceived risk levels. Using decision tree’s categorizing techniques to find the apparent factors that affect risk in fund portfolio; use genetic algorithm’s search and fitness-based process, to construct the fund portfolio based on investor’s needs. This research uses the beta-coefficient as the base of selection of mutual fund risk level. As for selection of the funds, the main factor is the Rate of Return, along with six overall fund performance indexes: standard deviation, beta-coefficient, sharpe index, Jensen index, treynor index and information ration, and seven fund characteristics that could affect the performance of the fund. For asset allocation, the percentage of any fund is based on the fund portfolio calculated using genetic algorithm. Our experimental concludes that the funds chosen with the combination of decision tree and generic algorithm perform better. Of the funds chosen, the R3 type fund portfolio has the more apparent effect.

參考文獻


47. 許慶聰,使用基因演算法完成綜合來源編碼與通道編碼設計之研究,中原大學電機工程研究所碩士論文,2001。
41. 施正宏,結合總體經濟指標及個股財報資料以預測個股漲跌-以台灣電子類股為例,中原大學資訊管理研究所碩士論文,2004。
46. 張愷芬,資料採礦-保險業商品市場分析,淡江大學資訊工程研究所碩士論文,2002。
4. Carhart, M. M., “On Persistence in Mutual Fund Performance,” The Journal of Finance, Vol.52, 1997, 57-81.
5. Chevalier, Judith and Glenn E., “Are Some Mutual Fund Managers Better than Others? Cross-sectional patterns in behavior and performance,” The Journal of Finance, Vol.54 (3), 1999, 878-899.

被引用紀錄


劉婷文(2007)。共同基金評比與績效之實證研究〔碩士論文,淡江大學〕。華藝線上圖書館。https://doi.org/10.6846/TKU.2007.00825
鍾孟杰(2014)。以資料探勘及多重分類器技術建構企業財務危機預警模型〔碩士論文,中原大學〕。華藝線上圖書館。https://doi.org/10.6840/cycu201400590
洪齊尉(2009)。整合遺傳演算法與粒子群最佳化演算法 於投資組合最佳化問題之研究〔碩士論文,國立臺北科技大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0006-2406200911233500

延伸閱讀