A New Application of Hidden Markov Model in Exchange Rate Forecasting

This paper presents a new application of Hidden Markov Model (HMM) as a forecasting tool for the prediction of the currency exchange rate between the US dollar and the euro. The results obtained show that the difference between price gaps which consists open, high, and low price can be selected to produce the best model parameter of Hidden Markov Model. Three model parameters based on Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC) and Log likelihood value has been calculated. The empirical results show that the projection from the daily opening price to the highest and lowest price of the day produce lowest AIC, BIC and Log likelihood values respectively. The price pattern based on the daily closing value has been ignored because of poor model performance compared to the other two price gaps. As it turns out, the proposed application outperforms the model which uses the closing price as an input variable in terms of model parameter based on AIC, BIC, and Log likelihood values. Keywords— Hidden Markov Model, Log likelihood, AIC, BIC, time series forecasting, exchange rates