MODELLING THE NAIRA/U.S. DOLLAR CURRENCY EXCHANGE RATES USING DECISION TREE, ORDINARY LEAST SQUARES AND RANDOM FOREST MACHINE LEARNING ALGORITHMS

U. C. IBEKWE; L. A. AJIJOLA

U. C. IBEKWE Department of Actuarial Science & Insurance Faculty of Management Sciences, University of Lagos. Nigeria.
L. A. AJIJOLA Department of Actuarial Science & Insurance Faculty of Management Sciences, University of Lagos. Nigeria.

Keywords: Currency Exchange Rates, Decision Tree, Machine Learning, OLS, Random Forest

Abstract

The exchange rate of a country’s currency is the fundamental price in any economy and provides a measure of the value of that country’s currency relative to other currencies. This paper examines the issues of model fit and prediction accuracy of three modelling techniques that could be employed in analyzing historical data regarding currency exchange rates of the Naira with the U.S. Dollar as a reference. The techniques of Ordinary Least Squares (OLS), Decision Trees (DTs), and Random Forest were analyzed and their performance was compared in terms of model fit and prediction accuracy. The study found that Random Forest offered the best results, followed by Ordinary Least Squares and then Decision Trees, in that order. The implication and recommendation, therefore, is that the Random Forest model should be preferred in future studies in this area, although OLS also gives reasonable results and could also be deployed.