Spatial disaggregation to generate city-scale travel demand models

Abstract

Many activity based models specify locations at the traffic analysis zone (TAZ) resolution. In city scale travel models and for MaaS predictions, a finer grained spatial resolution may be required. An artificial neural network was used to classify predicted daily schedules based on the total travel duration using a household travel survey. We propose a TAZ to street address based disaggregator that first generates a choice set of schedule variants and then selects the final candidate according to the schedule specific probability weight function delivered by the classifier coefficients. This paper describes how the technique has been applied to The Netherlands. It shows that realistic schedules are produced using a zoning having a large variety in TAZ size.