運用基因演算法買賣期貨之研究—以臺股期貨為例

The Application of Genetic Algorithms for Trading TAIEX Futures

徐弘翰(Hung-Han Hsu)

指導教授 ： 李富民

朝陽科技大學/資訊學院/資訊管理系/碩士(2014年)

摘要

本研究以臺股期貨為交易標的，組合及參數化數個常用的臺股現貨技術指標為交易規則，並運用基因演算法找出最佳的交易規則。為了比較以獲勝率為導向或以報酬為導向找出的最佳交易規則的績效，本研究設計了連續10個年份的實驗情境，包含盤整向上型、盤整向下型及多空分歧型。相較於以報酬為導向找出的最佳交易規則，實驗結果顯示以獲勝率為導向找出的最佳交易規則，絕大多數能獲得較高的獲勝率，且超過半數也能夠獲得較高的報酬，特別是長期(連續10個年份)的總合報酬表現。

關鍵詞：交易規則、基因演算法、獲勝率、技術指標、臺股期貨

關鍵字

技術指標 ； 獲勝率 ； 基因演算法 ； 交易規則 ； 臺股期貨

並列摘要

Technical trading rules for TAIEX futures are studied in this research. They are parameterized and composed of several conventional technical indicators of TAIEX. To find the optimal parameters of the rules, genetic algorithms are used for specific learning objectives. Ten experiment cases in a decade that including swing upward, swing downward, and uncertainty cases are designed to compare the performance of the winning-percentage-oriented or the profit-oriented optimal trading rules. The results show that the winning-percentage-oriented optimal trading rules are superior. Their winning percentage performance is better in most cases and the profit performance is better in more than half of cases, especially, the total profit performance in the 10 cases.

Keywords: Trading Rules, Genetic Algorithms, Winning-percentage, Technical Indicators, TAIEX Futures

並列關鍵字

TAIEX Futures ； Technical Indicators ； Winning-percentage ； Genetic Algorithms ； Trading Rules

參考文獻

[ 5] 林德祥(2010)，應用模糊時間序列與倒傳遞類神經網路預測台灣加權股票指數趨勢，碩士論文，國立交通大學工業工程與管理學系，新竹。

[11] 莊政文(2008)，以獲勝率為導向的最佳化技術交易規則—以台股期貨為交易標的，碩士論文，朝陽科技大學資訊管理系，臺中。

[15] 蔡伯煜(2011)，K 線圖探勘於股票預測之研究，碩士論文，國立中央大學資訊管理學系，桃園。

[19] G. S. Atsalakis and K. P. Valavanis (2009), “Forecasting Stock Market Short-term Trends Using a Neuro-fuzzy Based Methodology,” Expert Systems with Applications, Vol. 36, No. 7, pp. 10696-10707.

[21] P. C. Chang, C. H. Liu, J. L. Lin, C. Y. Fan, and C. P. Ng (2009), “A Neural Network with a Case based Dynamic Window for Stock Trading Prediction,” Expert Systems with Applications, Vol. 36, No. 3, pp. 6889-6898.