Estimating Parking Search Times for Transport Modelling in Europe: A Hierarchical Bayes Approach with Cross-Regional Data Integration

Abstract

Transport simulation frameworks often exclude parking because data are scarce. This study show how to predict parking search times as inputs for transport simulation frameworks for Europe, where only few observations are available, by using a prior distribution constructed from a large data set on parking search times in North America. An accelerated failure time model is estimated using Hierarchical Bayes methods so that information between parameters for different regions and data sources is shared. We make probabilistic predictions for search times in Zurich and Berlin. Our results show that an appropriate model needs to capture both the likelihood of finding parking as well as search time. Our results highlight the usefulness of drawing on different data sources for parameter estimation, and the importance of generalising from sample to population when simulating a range of plausible parking search times at a given location.