Evaluation of FOREX trading Strategies based in Random Forest and Support Vector Machines

The Foreign Exchange (Forex) is the largest market in the world and has a daily trading volume of approximately 3.2 trillion dollars. The price movements are influenced by many exogenous factors, thereby it is difficult to predict. As a result, modeling its movement would enable high profitable investment strategies. Aiming at setting up a model that helps the market practitioner make better decisions for trading on Forex, Machine Learning algorithms (Random Forest and SVM) are adopted in this work. Classic technical indicators, like moving averages, are used as features for these algorithms. In order to evaluate these approaches, several simulations were carried out on the pairs Euro/Dollar, Pound/Dollar, Dollar/Swiss Franc and Dollar/Japanese Yen, using three different metrics. For virtually all scenarios investigated, the proposed algorithms outperform the traditional technical indicators.