Estimation of dollar rate changes in Turkey using Hidden Markov models

Abstract

In this work, a prudential estimation method for U.S. dollar rate, as the exchange rate, is presented. As the estimation method, Hidden Markov Models, which are described as Markov Chains with some additional properties, are used. In the model, estimations for dollar rate change for the year 2008 are made using dollar rate values and related economic data observed in Turkey between the years 1992 - 2007. In the Hidden Markov Models, mathematical calculations belonging to the model are very complicated, thus require pretty much time to be solved. For this reason, in the calculations, Hidden Markov Model solution algorithms inside the software Matlab2007 will be used. Two estimations for the year 2008 are made, where the first one is for January and February and the second one is for January, February and March. The accuracy of the estimated dollar rate changes came out to be fairly high.