Comparing models to forecast cargo volume at port terminals

Abstract
Economic growth has a direct link with the volume of cargo at port terminals. To encourage growth, investment decisions on infrastructure are required that can be performed by the development of econometric models. We compare three time-series models and one machine-learning model to estimate and forecast cargo volume. We apply an ARIMA+GARCH+Bootstrap, a multiplicative Holt-Winters, a support vector regression model, and a time-series model with explanatory variables ARIMAX. The models forecast cargo through the ports of San Pedro using data from 2008 to 2016. The database contains imports and exports of bulk, container, reefer, and ro-ro cargo. Results show that the multiplicative Holt-Winters model is the best method to forecast imports and exports of bulk cargo, while the support vector regression model is the best method to forecast imports and exports of container, reefer, and ro-ro cargo. The Diebold-Mariano Test, the RMSE metric, and the MAPE metric validate the results.