集成基因表达规划法应用于动态股票交易策略探勘之研究

The Mining of Dynamic Stock Trading Strategies Based on Ensemble Gene Expression Programming

本研究主要目的乃运用基因表达规划法(Gene Expression Programming,GEP),加入集成学习(Ensemble Learning)与权重机制,以及设计动态时间天期技术指针的机制,期望提升程序交易(Program Trading)领域中交易策略设计的弹性与效果。因此,以台湾三家股票上市公司为实验目标,期望检验这个智能型的集成基因表达规划法(Ensemble-GEP,E-GEP)的交易策略系统,是否能帮助投资人发展出好的交易策略,提升风险报酬比率。根据实验结果:(1)使用动态天期技术指针的交易策略确实优于固定天期的交易策略,更能捕捉到交易讯号,获致较高的夏普率(Sharpe Ratio)与投资报酬率;(2)E-GEP交易系统获得较高的夏普率,优于使用单一分类器的模型(SGEP),显示其能提升风险报酬比率;本研究研发这套创新的交易策略系统,可建构在MultiCharts.NET程序交易平台,提供一个学术与实务应用结合的具体典范。

Genetic expression programming,ensemble learning,weights and dynamic technical indexes are used in the research.expceting to enhance the flexibility and performance of trading strategic design in the field of program trading.Therefore,three exchange listed companies in Taiwan are used as experimental subject,expecting to examine the trading strategic model of innovative Ensemble-Gene Expression Programming(E-GEP)and help investors find the suitable trading strategies to increase the return of investment in the stock market.The results of the experiments Show that:(1)compared with the normal technical indexes,the dynamic technical indexes are better to find the appropriate trading strategies and get higher Sharpe ratio and return of investment;(2)using ensemble method are the necessary complement to using single classifier,and the former can get the higher Sharpe ratio.