In this paper, a preliminary investigation of Cartesian Genetic Programming (CGP) for algorithmic intraday trading is conducted. CGP is a recent new variant of genetic programming that differs from traditional approaches in a number of ways, including being able to evolve programs with limited size and with multiple outputs. CGP is used to evolve a predictor for intraday price movements, and trading strategies using the evolved predictors are evaluated along three dimensions (return, maximum drawdown and recovery factor) and against four different financial datasets (the Euro/US dollar exchange rate and the Dow Jones Industrial Average during periods from 2006 and 2010). We show that CGP is capable in many instances of evolving programs that, when used as trading strategies, lead to modest positive returns.
|Cite as: Mayo, M. (2012). Cartesian Genetic Programming for Trading: A Preliminary Investigation. In Proc. Data Mining and Analytics 2012 (AusDM 2012) Sydney, Australia. CRPIT, 134. Zhao, Y., Li, J. , Kennedy, P.J. and Christen, P. Eds., ACS. 149 - 156 |
(local if available)