Hidden Markov Models for Forex Trends Prediction

Abstract

Foreign Exchange (Forex) market is a complex and challenging task for prediction due to uncertainty movement of exchange rate. However, these movements over timeframe also known as historical Forex data that offered a generic repeated trend patterns. This paper uses the features extracted from trend patterns to model and predict the next day trend. Hidden Markov Models (HMMs) is applied to learn the historical trend patterns, and use to predict the next day movement trends. We use the 2011 Forex historical data of Australian Dollar (AUS) and European Union Dollar (EUD) against the United State Dollar (USD) for modeling, and the 2012 and 2013 Forex historical data for validating the proposed model. The experimental results show outperforms prediction result for both years.