Hybrid ARIMA and Machine Learning Approach for the VND/USD Exchange Rate Prediction in Vietnam: A Comparative Analysis

The exchange rate plays a central role in macroeconomic stability in Vietnam. This work makes an effort to predict the VND/USD exchange rate by hybrid approaches between the Autoregressive integrated moving average (ARIMA) and two machine learning algorithms, namely Random forest (RF) and Artificial neural network (ANN). Using daily data between Jan 3, 2000, and July 11, 2023, we find that the machine learning approaches, i.e., ANN and RF, are better than ARIMA in forecasting the VND/USD. Moreover, the hybrid approaches ARIMA-ANN and ARIMA-RF utilize the linear component by ARIMA and the non-linear part by ANN/RF to minimize forecasted errors. The results show the predictive improvement compared to ARIMA or ANN/RF single models. The RF and its hybrid ARIMA-RF have outperformed the others regarding forecasting tasks.