Multi-scale Foreign Exchange Rates Ensemble for Classification of Trends in Forex Market

Abstract

Foreign exchange (Forex) market is the largest trading market in the world. Predicting the trend of the market and performing automated trading are important for investors. Recently, machine learning techniques have emerged as a powerful trend to predict foreign exchange (FX) rates. In this paper, we propose a new classification method for identifying up, down, and sideways trends in Forex market foreign exchange rates. A multi-scale feature extraction approached is used for training multiple classifiers for each trend. Bayesian voting is used to find the ensemble of classifiers for each trend. Performance of the system is validated using different metrics. The results show superiority of ensemble classifier over individual ones.