Comparison of Two Approaches for Modeling Freight Movement at Seaports

Abstract

Seaports host international cargo operations and are primary generators of freight traffic in the United States. Truck/rail trip generation and modal split models provide public agencies with valuable information necessary to prioritize funds for roadway upgrade projects and port infrastructure modifications. This paper presents two approaches for developing freight trip generation models: regression analysis and backpropagation neural networks (BPN). These models are used for predicting the levels of cargo truck traffic moving inbound and outbound at seaports. Based on the Port of Miami case, it was found that the BPN model is more accurate than the regression model. However, the BPN model requires a sizable database. Using the BPN approach, the paper presents a new combined truck trip generation and truck-rail modal split model for the Port of Jacksonville. It was found that the primary factors affecting truck-rail volume are the amount and direction of cargo vessel freight, commodity type, and the particular weekday of operation. In summary, the neural network model results were found significantly accurate for both Florida ports.