Artificial Intelligence in International Finance An Empirical Comparison of Traditional and Machine-Learning Models for Brent Crude Oil Price Forecasting
0 viewsAbstract
Oil prices are important to international finance because they affect the cost of imports, inflation, company earnings and, in many countries, the balance of payments. They are also difficult to predict because the market reacts to changes in supply, demand, financial conditions and unexpected events. This paper compares four forecasting approaches for monthly Brent crude oil prices: a random-walk benchmark, an ARIMA model, a Random Forest model and a multilayer perceptron (MLP) neural network. The data cover January 2010 to December 2025. The first 80 percent of the observations are used for model estimation and the remaining 20 percent are kept for out-of-sample testing. The results show clear differences between the models. The Random Forest gives the smallest MAE and RMSE in the test period, while the MLP records the highest directional accuracy. ARIMA also performs better than the random-walk benchmark on the main error measures. The results suggest that machine-learning methods can be useful for forecasting oil prices, but their performance depends on the measure used to judge the forecast. For international finance, this is important because a model that gives a small numerical price error is not necessarily the same model that gives the most accurate prediction of whether prices will rise or fall.
Keywords
ARIMA, Artificial Intelligence, Brent Crude Oil, International Finance, Machine Learning, Neural Networks, Price Forecasting, Random Forest