Dynamic trading decision support system using rule selector based on genetic algorithms

Abstract

We propose a dynamic trading decision support system (DTDSS) capable of selecting a near optimal rule combination for each time interval. The system provides buying, holding and selling signals from which a decision can be made based on which signal exceeds a predetermined threshold. These signals are obtained by extracting features from various stock indices using rule inference network based on AND/OR graphs. We show that simply applying an identical rule combination to all time intervals is insufficient in making quality decision. In DTDSS, an intelligent rule selector (GARS) using genetic algorithms and a moving window scheme are combined to determine an optimal rule combination for each different time interval. Experimental results on Taiwan stock exchange weighted stock index (TSEWSI) show that DTDSS outperforms the simple "buy and hold" strategy. Trading decision support systems with OARS are shown to yield much more profits than otherwise without OARS.