Abstract: Traders in security market generally create and adapt a trading strategy pool first and choose one to execute according to market states reciprocally. In trading strategy layer, traditional Constant Proportion Portfolio Insurance (CPPI) strategy only considers protecting floor wealth. This study considers a goal-directed strategy to protect goal wealth in advance. Combining the CPPI strategy and the goal-directed strategy, this research adopts a piecewise linear goal-directed CPPI (GDCPPI) strategy to trade securities. In technology layer, this research applies the Learning Classifier System (LCS) to form and adapt the trading strategy pool and choose a suitable trading strategy against the consecutively changed market states. With the help of learning classifier system, this study can deal with the dynamic pattern evolution of the piecewise linear goal-directed CPPI strategy. This study executes many experiments under Brownian motion, GA and LCS technologies to generate the piecewise linear goal-directed CPPI strategies. Experimental results show that the LCS technology outperforms GA technology and GA technology outperforms Brownian motion technology further in generating piecewise linear goal-directed CPPI strategies.