An Investigation of Temporal Transferability Comparing Trip-Based and Tour-Based Choice Models

Transferability measures how well a model developed in a specific temporal, spatial, transportation and institutional context can be applied to another context. This paper presents a study in examining the temporal transferability of trip generation choice models. Two types of trip purposes are investigated, commute and shopping, to investigate whether the trip-making for one purpose tends to be more stable over time than that for the other purpose. Presumably, commute behavior may have higher chances of stability as it relates to work activities that tend to be less likely subject to other influences. Moreover, this study also investigated two modeling approaches, trip-based and tour-based, to examine whether one approach performs better than the other in terms of temporal transferability. Trip generation choice models are developed based on two waves of the Regional Household Travel Survey (RHTS) for the New York metropolitan region, conducted in year 1997-1998 and year 2010-2011, respectively. Various transferability tests were conducted at both aggregate and disaggregate levels. Results indicate that at aggregate level shopping models perform slightly better than the commute models, for both trip-based and tour-based approaches. Similarly, tour-based approach shows slightly better performance than trip-based approach at aggregate level. Overall all models fail to exhibit significant temporal stability.