The Application of the Genetic Algorithm in Promoting Stock Trading Performances

Abstract

This thesis joins the debate on utilizing the Genetic Algorithm (GA) to discover profitable trading strategies by providing an out-of-sample test of GA-based trading strategies on the CSI 300 index. Our results suggest that, with trading costs taken into consideration, GA-based trading rules consistently beat the buy-and-hold strategy in daily trading of CSI 300 index. Besides, we open up the black box of the evolution process of the GA by testing the statistical significance of the GA-based profitable trading strategies through the Fama-MacBeth regressions. In addition, this study connects the literature on the regime switching with studies on the GA-based trading strategies to construct one regime-switching Genetic Algorithm (RSGA) model and makes a comparison between the GA-based and the RSGA-based trading strategies. The empirical results show that trading strategies generated from the RSGA model consistently outperform those obtained from the GA model.