Leveraging Machine Learning for Exchange Rate Prediction: Fresh Insights from BRICS Economies

Abstract

In the background of de dollarisation and continuous uncertainty looming around oil prices, this research assesses how well machine learning-based linear regression models predict currency exchange rates for the BRICS nations using oil prices as a predictive factor. The analysis focuses on evaluating the effectiveness of these models in forecasting exchange rates for these five emerging economies. This study shows that oil prices serve as key indicators for all five currencies examined. The models’ performance was evaluated using three statistical measures: MSE, RMSE, and R-Squared. The results suggest that, among the currencies studied, the Chinese Yuan demonstrated the most accurate predictions, as evidenced by its lowest MSE and RMSE values. By contrast, the South African Rand displayed the highest R-squared value, indicating that oil prices had a greater explanatory capacity for exchange rate fluctuations. Nevertheless, the models demonstrate limited predictive power, indicating a disconnection among oil prices and forex rates in these nations. The findings suggest that relying solely on oil prices may not provide accurate exchange rate predictions and that considering additional variables could improve the effectiveness of the models.