Bayesian approach to model choice analysis in freight transport models (Case study: Central Bioceanic Railway Corridor)

Abstract

Transport planning requires tool to model the current and future situation of an infrastructures network. In this way, different scenarios of passenger flows, vehicles or freight can be predicted and serve as information for decision making. One of these tools are the so called "Demand models", among which the four steps models (Generation/attraction, Distribution, Modal choice, Network assignment) is a remarkable example for its widespread use. This paper presents a novel Bayesian approach to the third step of a demand transport model. Traditional discrete choice models are the ones most commonly used at this purpose, although other methods such as neural networks have been used by some authors. A Bayesian network is proposed as tool for estimating the decisions made by users when they face the need to choose which transport alternatives to use for sending cargo in a case study corresponding to the Central Bioceanic Corridor in South America. The results from fitting a logit model and a Bayesian network are compared and show the Bayesian network to be a promising tool to be applied in this kind of applications.