Computational learning techniques for intraday FX trading using popular technical indicators

There is reliable evidence that technical analysis, as used by traders in t he foreign exchange (FX) markets, has predictive value regarding future mov ements of foreign exchange prices. Although the use of artificial intellige nce (AI)-based trading algorithms has been an active research area over the last decade, there have been relatively few applications to intraday forei gn exchange-the trading frequency at which technical analysis is most commo nly used. Previous academic studies have concentrated on testing popular tr ading rules in isolation or have used a genetic algorithm approach to const ruct new rules in an attempt to make positive out-of-sample profits after t ransaction costs. In this paper we consider strategies which use a collecti on of popular technical indicators as input and seek a profitable trading r ule defined in terms of them. We consider two popular computational learnin g approaches, reinforcement learning and genetic programming (GP), and comp are them to a pair of simpler methods: the exact solution of an appropriate Markov decision problem and a simple heuristic. We find that although all methods are able to generate significant in-sample and out-of-sample profit s when transaction costs are zero, the genetic algorithm approach is superi or for nonzero transaction costs, although none of the methods produce sign ificant profits at realistic transaction costs,We also find that there is a substantial danger of overfitting if in-sample learning is not constrained .