First Advisor

Liming Wang

Term of Graduation

Spring 2022

Date of Publication


Document Type


Degree Name

Doctor of Philosophy (Ph.D.) in Urban Studies: Regional Science


Urban Studies and Planning




Automated vehicles, Choice of transportation -- Mathematical models, Machine learning



Physical Description

1 online resource (xii, 194 pages)


New mobility technologies, such as shared mobility services (e.g., car-sharing) and, more importantly, autonomous vehicles (AVs), continue to evolve. The supply-side advancement will likely disrupt and transform transportation mode choice behaviors, and create a new paradigm since they are emerging and becoming increasingly feasible alternatives to the existing modes of transportation. Accordingly, this dissertation employs discrete choice modeling (DCM) and machine learning (ML) using a U.S. nationwide stated choice experiment to understand how travelers adopt new transportation modes or continue to use conventional modes of transportation.

This dissertation consists of three papers. The first examines future market shares of each available mode of transportation in the era of AVs, factors influencing mode choice behaviors, and their marginal effects using a mixed logit model. The second uses interpretable ML to investigate the optimal algorithm (i.e., stochastic gradient boosting decision tree model) in greater depth, including feature importance and non-linear marginal effects. Focusing on methodology, the final paper assesses the limitations of ML when applied to transportation mode choice modeling and suggests future research directions for methodological improvements by comparing ML to DCM.

The dissertation contributes to three major elements of the current understanding of transportation mode choice behavior in the era of AVs and choice modeling as follows: First, consumers in the AV era could choose from a variety of transportation modes likely to coexist, including private AVs, shared mobility services, and conventional transportation modes. This dissertation thus makes a significant contribution by examining more comprehensive transportation mode choice behaviors and expanding demand-side discussions. Second, since current transportation planning efforts have relied on estimates and expectations, this dissertation contributes to the decision-making process by offering crucial underlying knowledge not currently available. Third, this dissertation assesses the limitations of ML for transportation mode choice modeling and suggests potential future avenues for methodological improvement.


© 2022 Sangwan Lee

In Copyright. URI: This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s).

Persistent Identifier