A Model Typology Approach for Improving Spatial Transferability of Freight Demand Estimates: Case Study of Geographically Dissimilar Regions in India

Spatial transferability of freight demand models is a practice that is needed to enable the use of formerly developed models in estimating the freight demand of another region. This practice is particularly crucial for planning agencies in densely populated countries like India, needing more resources for comprehensive freight data collection. There are studies on the transferability of freight demand models. However, the transferability of non-parametric freight demand models has yet to be investigated. It is necessary to understand whether the extent of transferability of non-parametric models is greater than that of parametric models so that planning agencies can adopt more reliable modelling approaches/typologies. This paper investigates the spatial transferability of freight production models using parametric (ordinary least squares regression and robust regression) and non-parametric (multiple classification analysis and support vector regression) modelling approaches. The transferability assessment results show that the non-parametric models are more transferable than non-parametric models.