Nishant Kumar

Nishant Kumar

Postdoctoral Research Fellow at NUS; transport AI, congestion prediction, and interpretable ML.

Code & Data

Looking for reusable tools and packages? See the Software page.

In my 2021 survey on deep learning for traffic congestion detection, prediction, and alleviation, I emphasized a recurring challenge in transportation AI: the lack of open datasets, open-source implementations, and reproducible benchmarks.

Since then, I have tried to make openness a consistent part of my own research practice. For papers where I am a first author or co-author, I aim to share the associated code whenever possible. I also try to use freely available datasets when they are suitable for the research question, so that the work can be reproduced, compared, extended, and used by a wider community.

This page collects paper-specific code repositories, dataset links, reproducibility notes, and related resources.

Paper-Specific Repositories

Enhancing Deep Learning-Based City-Wide Traffic Prediction Pipelines Through Complexity Analysis

Year: 2024
Role: First author
Code: GitHub link
Data: Dataset link / freely available data
Paper: DOI link

Brief description of what the repository contains.

Uncertainty Quantification for Image-Based Traffic Prediction Across Cities

Year: 2023
Role: Co-author
Code: GitHub link, if available
Data: Dataset link, if available
Paper: arXiv / DOI link

Brief reproducibility note.

Quantifying the Impacts of Non-Recurrent Congestion on Workplace EV Charging Infrastructures

Year: 2025
Role: First author
Code: GitHub link, if available
Data: Dataset link, if available
Paper: DOI link

Brief reproducibility note.

Openness and Reproducibility

I try to follow three principles when sharing research outputs:

  1. Release paper-specific code whenever possible.
  2. Use freely available datasets when they are suitable for the research question.
  3. Document enough of the workflow so that others can reproduce, compare, or extend the work.

Some projects may not include full code or data because of licensing, collaborator restrictions, proprietary inputs, or privacy constraints. In such cases, I try to provide as much information as possible about the methodology, data sources, and reproducibility limitations.