An Evolutionary Approach for Forex Trading Based on Technical Indicators

Alexandre Delbem (ICMC-USP)

Abstract: We propose an evolutionary algorithm to determine the parameters of rules of a trading system applied to Foreign Exchange Market. This system is composed by a set of trading rules based on technical indicators and it aims at identifying when to open, close and manage orders. Each individual is a set of possible values for the parameters to be optimized. The evaluation of the population uses a given time series of a currency pair in order to calculate the profitability of each individual. Later the solutions with the best fitness are validated against a time series succeeding the aforementioned and the results are discussed. Different indicator ratios are used to assess the solutions’ ability of achieving profit. Results show that the trading rules generated using the proposed algorithm can produce reliable market predictions, but their performance is reduced when transaction costs are taken into account. (Joint work with Geraldo Silva and Victor Hugo de Oliveira).