Skip to content ↓

SMART breakthrough uses artificial neural networks to enhance travel behavior research

Theory-based residual neural network combines discrete choice models and deep neural networks, long viewed as conflicting methods.
Press Inquiries

Press Contact:

Linda Zahka-Stockdale
Phone: 617-253-0522
Singapore-MIT Alliance for Research and Technology
Photo of a busy city intersection, with vehicles backed up trying to make a turn onto a high-traffic street
“Improved insights to how travelers make decisions about travel mode, destination, departure time, and planning of activities are crucial to urban transport planning for governments and transport companies worldwide,” says MIT postdoc Shenhao Wang.

Researchers at the Future Urban Mobility (FM) interdisciplinary research group at Singapore-MIT Alliance for Research and Technology (SMART), MIT’s research enterprise in Singapore, have created a synthetic framework known as theory-based residual neural network (TB-ResNet), which combines discrete choice models (DCMs) and deep neural networks (DNNs), also known as deep learning, to improve individual decision-making analysis used in travel behavior research.

In their paper, "Theory-based residual neural networks: A synergy of discrete choice models and deep neural networks," recently published in the journal Transportation Research: Part B, SMART researchers explain their developed TB-ResNet framework and demonstrate the strength of combining the DCMs and DNNs methods, proving that they are highly complementary.

As machine learning is increasingly used in the field of transportation, the two disparate research concepts, DCMs and DNNs, have long been viewed as conflicting methods of research.

By synergizing these two important research paradigms, TB-ResNet takes advantage of DCMs’ simplicity and DNNs’ expressive power to generate richer findings and more accurate predictions for individual decision-making analysis, which is important for improved travel behavior research. The developed TB-ResNet framework is more predictive, interpretable, and robust than DCMs or DNNs, with findings consistent over a wide range of datasets.

Accurate and efficient analysis of individual decision-making in the everyday context is critical for mobility companies, governments, and policymakers seeking to optimize transport networks and tackle transport challenges, especially in cities. TB-ResNet will eliminate existing difficulties faced in DCMs and DNNs and allow stakeholders to take a holistic, unified view toward transport planning.

Urban Mobility Lab at MIT postdoc and lead author Shenhao Wang says, “Improved insights to how travelers make decisions about travel mode, destination, departure time, and planning of activities are crucial to urban transport planning for governments and transport companies worldwide. I look forward to further developing TB-ResNet and its applications for transport planning now that it has been acknowledged by the transport research community.”

SMART FM lead principal investigator and MIT Department of Urban Studies and Planning Associate Professor Jinhua Zhao says, “Our Future Urban Mobility research team focuses on developing new paradigms and innovating future urban mobility systems in and beyond Singapore. This new TB-ResNet framework is an important milestone that could enrich our investigations for impacts of decision-making models for urban development.”

The TB-ResNet can also be widely applied to understand individual decision-making cases as illustrated in this research, whether it is about travel, consumption, or voting, among many others.

The TB-ResNet framework was tested in three instances in this study. First, researchers used it to predict travel mode decisions between transit, driving, autonomous vehicles, walking, and cycling, which are major travel modes in an urban setting. Secondly, they evaluated risk alternatives and preferences when monetary payoffs with uncertainty are involved. Examples of such situations include insurance, financial investment, and voting decisions.

Finally, they examined temporal alternatives, measuring the tradeoff between current and future money payoffs. A typical example of when such decisions are made would be in transport development, where shareholders analyze infrastructure investment with large down payments and long-term benefits.

This research is carried out by SMART and supported by the National Research Foundation (NRF) Singapore under its Campus for Research Excellence And Technological Enterprise (CREATE) program.

The Future Urban Mobility research group harnesses new technological and institutional innovations to create the next generation of urban mobility systems to increase accessibility, equity, safety, and environmental performance for the citizens and businesses of Singapore and other metropolitan areas worldwide. FM is supported by the NRF Singapore and situated in CREATE.

SMART was established by MIT in partnership with the NRF Singapore in 2007. SMART serves as an intellectual and innovation hub for research interactions between MIT and Singapore, undertaking cutting-edge research projects in areas of interest to both Singapore and MIT. SMART currently comprises an Innovation Center and five interdisciplinary research groups: Antimicrobial Resistance, Critical Analytics for Manufacturing Personalized-Medicine, Disruptive and Sustainable Technologies for Agricultural Precision, FM, and Low Energy Electronic Systems.

Related Links

Related Topics

Related Articles

More MIT News