TANKER FREIGHT FORECASTING: ARIMA VERSUS ARTIFICIAL NEURAL NETWORKS (ANNs), A COMPARISON

This article addresses time series forecasting in the context of maritime transportation of crude oil. More specifically, it compares two approaches for projecting monthly tanker freight rates: ARIMA and Artificial Neural Networks (ANNs). The results of the accuracy metrics for 12-month out-of-sample forecasts indicate that both approaches are reliable and useful, however, ANNs outperformed ARIMA across all metrics, achieving a MAPE of 6.84% versus 13.01%, an MAE of 75.18 compared to 125.39, and an RMSE of 109.38 versus 136.02. Finally, the article recommends future research of ANNs in this context, as well as hybrid approaches.