Abstract
This study examines the transferability of tour-generation models between three metropolitan regions in Florida. Naïve transfer methods are examined to assess the performance of the transferred models (from two other regions) to that of the locally estimated model. The assessment is done in the context of the generation of four tour purposes. Transferability is evaluated with multiple measures such as aggregate and disaggregate predictive abilities and the aggregate elasticities to specific socioeconomic factors. The results indicate that a transferred model that does best with aggregate predictions is not guaranteed to give better performance with elasticities regarding specific factors. For any pair of regions, and for a given metric for assessing transferability, the models for all tour purposes are not equally transferable. Finally, transfer ability is not symmetric. All these issues underscore the need to administer caution in borrowing parameters from one region for use in another.
Get full access to this article
View all access options for this article.