Disaggregate Probabilistic Models to Predict Trip Generation Propensities and Mode Choice Behaviour

Abstract

Travel behaviour models serve as important tools to understand what factors affect trip generation and preferred mode of travel in different contexts. These models can be used to quantify the impacts and assess the consequences of development plans and policy actions. Growing populations and the increase in travel demand warrant investigation into determinants of travel behaviour that can guide policy changes and infrastructure investments. This thesis uses two cross-sectional datasets containing various sociodemographic and land-use attributes from 2015 and 2019. Trip-generation propensities for transit and automobile modes are predicted using a bivariate ordered probit approach which enables the determining of factors affecting the trip-generation propensity of each mode while establishing the correlation between their propensities. Further, a multinomial logit model is estimated to investigate the determinants of mode choice for home-based discretionary trips. The results show that improvement in areas with low transit accessibility can considerably increase the transit trip-making propensity and living in downtown Toronto increases the probability of choosing active modes of travel.