A Comparison of Forecasting Accuracy of Machine Learning Models: A Study on Selected Foreign Exchange Rates
Exchange rate, an economic indicator of the country is the relative price of one country’s currency in terms of another country’s currency. Stability in exchange rate is important for stable economic growth. Exchange rates, a financial time series highly fluctuate and are chaotic in nature. Forecasting exchange rate fluctuations is very important to countries’ economy. Many researchers reported that the machine learning models, namely Artificial Neural Network (ANN) models and Support Vector Regression (SVR) models provided good forecasts. The aim of this research was to compare the forecasting accuracy of ANN and SVR models in forecasting daily exchange rates of Sri Lankan Rupees to Euro and Yen and to identify a better model. Daily time series data, collected from 2nd July, 2012 to 31st August, 2016 (1008 trading days) from the official website of Central Bank of Sri Lanka, were analysed using Eviews, MATLAB and R packages. A Nonlinear autoregressive neural network (NAR) model using Scaled Conjugate Gradient (SCG) learning algorithm, a SVR model with Gaussian radial basis kernel function designed to model the exchange rate returns and a hybrid ANN with Intrinsic Mode Function (IMF) as input were designed to model the exchange rate series. Mean Squares Errors and directional accuracy measures revealed that the machine learning models explained the variation in the series well. However, Hybrid ANN models provided a more accurate directional accuracy as well as value forecast than NAR model and SVR model. These findings could be useful for domestic as well as foreign investors.
Exchange rates, Stochastic models, ANN, SVR, Hybrid ANN, IMFs, EMD
|University of Colombo
School of Computing
Managed & Published
This journal is published under a Creative Commons Attribution 4.0 International License.