Empirical Method for Predicting Internal-External Truck Trips at a Major Port

Abstract

This paper presents a case study to explore the truck-trip generation model for hauling containers at a major international seaport. An internal-external truck-trip forecast model is examined. It incorporates influential factors of regional freight activity attributes, economic growth attributes, and natural disaster attributes based on monthly data from 2000–2008. A best-fit truck-trip forecasting model is determined by comparing the prediction accuracy of a multiple regression model, time-series models, and a neural network model. The findings indicate that the back propagation neural network model generates better forecasting performance than the regression and time-series approaches. Additionally, this paper identifies the difference between truck trips and commodity-flow tonnages converted by truck payload factors, which would be significantly affected by truck-trip chains and truck drivers’ route choice behaviors. The analysis also reveals that a port truck-trip forecast model based on commodity flows would be very sensitive to the events of oil price fluctuations and new operation or infrastructure upgrade of competitive ports nearby, once the conversion difference goes up to 30%.