Description:

Dr. Prateek Bansal
Assistant Professor 
National University of Singapore (NUS)

Abstract:  
 
Improving interpretability at the minimum loss of flexibility/predictability is the main challenge in data-driven discrete choice models (DCMs). To this end, this study proposes a flexible and partially monotonic DCM by specifying the systematic utility using the Lattice networks (i.e., DCM-LN), which ensures monotonicity of the utility function relative to the selected attributed while learning the non-linear and interaction effects of attributes in a data-driven manner. DCM-LN estimates the attribute-specific non-linear effects as piecewise linear functions and considers their interactions using multilinear interpolation. The light architecture and an automated process to write monotonicity constraints make DCM-LN scalable and translatable to practice. The proposed DCM-LN is benchmarked against deep neural network-based DCM (i.e., DCM-DNN) and a parametric DCM in a simulation and empirical study.  With superior interpretability and high predictability, DCM-LN lays out new pathways to harmonize the theory-driven and data-driven paradigms. 
 
Bio: 

Dr. Prateek Bansal is a Presidential Young (Assistant) Professor at the National University of Singapore (NUS). Before joining NUS, he was a Leverhulme Trust Early Career Fellow at Imperial College London. Prateek leads the Behavioural and Cognitive Science Lab at NUS and is a co-principal investigator of the Adaptive Mobility, Land Use, and Infrastructure module at Future Cities Laboratory, Singapore. His research group is interested in creating methodological innovations at the intersection of Bayesian Machine Learning, Econometrics, Causal Inference, and Computational Psychology to address challenging questions related to mobility behaviour and the adoption of emerging technologies at an individual level and an urban scale. He is an Associate Editor of the Journal of Transport Economics & Policy and the Journal of Public Transportation and is a handling editor of Transportation Research Part B: Methodological. Apart from top Transportation journals, he regularly publishes in interdisciplinary journals like Energy Economics, Computational Statistics, and Statistics and Computing.