Investigating the Transferability of Individual Trip Rates: Decision Tree Approach

Transferring trip rates to areas without local survey data is a common practice which is typically performed in an ad-hoc fashion using household-based cross-classification tables. This paper applies a rule-based method called decision tree to develop individual-level trip generation models for eight different trip purposes as defined in the National Household Travel Survey data (NHTS 2009) in addition to their daily vehicle miles traveled (VMT). For each trip purpose, the models are then obtained by finding the best-fitted statistical distribution to each one of the final decision tree clusters while considering the correlation between different trip purposes. The rule-based models utilize several socio-demographic and land-use explanatory variables and are sensitive to changes in demographics. The performance of the models are then tested and validated in a transferability application to Phoenix Metropolitan Region. These models can be employed in a disaggregate microsimulation framework to generate trips with different purposes at individual or household level. They can also be used as an alternative solution for trip generation step of a conventional four step travel demand model.