組合編碼遺傳演算法於投資策略資金分配之應用; Developing Investment Strategy Portfolios by Combination Genetic Algorithms

NCU Institutional Repository > 管理學院 > 資訊管理研究所 > 博碩士論文 > Item 987654321/13248

請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/13248

題名:	組合編碼遺傳演算法於投資策略資金分配之應用;Developing Investment Strategy Portfolios by Combination Genetic Algorithms
作者:	吳詩敏;Shih-Min Wu
貢獻者:	資訊管理研究所
關鍵詞:	遺傳演算法;資金分配;擇時;投資組合;選股;capital allocation strategy;portfolio;genetic algorithms;stock selection strategy;market timing strategy
日期:	2006-05-23
上傳時間:	2009-09-22 15:27:12 (UTC+8)
出版者:	國立中央大學圖書館
摘要:	完整的投資決策問題應同時考慮到選股、擇時以及資金分配等三個策略，但此三個策略之間存在著複雜的互動關係，因此，在過去的研究中，通常分別針對三者做獨立的研究，藉以簡化問題的複雜度。　　本研究提出一個能同時整合選股、擇時及資金分配策略的架構，此架構的概念是延伸至傳統投資組合，不同於傳統投資組合事先決定一個擇時策略，來判斷投資組合內所有投資標的物進出場時機，本研究所提出之投資策略組合是透過排序篩選的步驟，同時決定投資標的物以及各標的物所適用的擇時策略，因此，於投資策略組合中，包含了該時間點適合投資的標的物以及各標的物於當時適用的擇時策略，不同標的物皆依據其適用的擇時策略來決定進出場時間點，不再侷限於同一個擇時策略。此外，本研究架構利用組合編碼遺傳演算法於組合問題上的最佳化搜尋能力，來決定資金分配的比重，以期能夠在投資過程中獲到最大報酬。　　本研究架構細分為「簡單投資策略組合」及「進階投資策略組合」，經由實驗證實，於2003年至2005年期間，簡單及進階投資策略組合在投資績效的表現上皆有突出的成效，顯示透過系統架構中的排序過程篩選投資策略，以及利用組合編碼遺傳演算法分配有限資金，皆能有效提升投資報酬。除此之外，本研究針對簡單及進階投資策略組合兩組實驗進行改良，加入作空的操作，以及事先採用前測期預測最佳排名區段。其結果顯示作空的操作一般而言能提升投資續效，但於多頭的情況下，卻難以準確掌握到作空進場時間點，反而會影響整體報酬。而事先採用前測期預測最佳排名區段，則能有效淘汰已過於吻合市場的投資策略，進一步加強了本研究架構之獲利能力。　　When it comes to the research of investing decision, three strategies should be considered at the same time, including stock selection strategies, market timing strategies, and capital allocation strategies. However, complicated interactive relations exist among the three strategies. In the past, the three strategies were studied independently to simplify the question. 　　The presented study proposes a framework, Investing Strategy Portfolio, combining stock selection strategies, market timing strategies, and capital allocation strategies. The concept of the framework is based on that of traditional portfolio investing. Nevertheless, unlike traditional portfolio investing which only one market timing strategy is used to decide when to buy or sell all stocks, this framework propose a strategy of ranking procedure which can help to decide what stocks should be picked up and when they should be bought/sold at the same time. Therefore, different market timing strategies rather than a unitary market timing strategy can be used. In addition, we weigh the allocation of capital to get most profit by Combination Genetic Algorithms in search optimization. 　　This present framework is divided into two types, including Simple Portfolio Investing Strategy and Advanced Portfolio Investing Strategy. Moreover, we use the data from 2003 to 2005. The results of the experiments show that both simple portfolio investing strategy and advanced portfolio investing strategy have excellent performance; however, tactics of going long and going short could influence the performance since it is difficulty to decide when to go short in bull market. Another discovery is that we can eliminate some overfitting investing strategy to improve the profitability of the present framework.
顯示於類別:	[資訊管理研究所] 博碩士論文

文件中的檔案:

檔案	大小	格式	瀏覽次數

在NCUIR中所有的資料項目都受到原著作權保護.

社群 sharing

資料載入中.....