Towards Scalable, Robust, and Generalizable Traffic Forecasting: A Data-Centric Approach
Confirmation seminar presented by Taeyoung Yu
Accurate traffic forecasting is critical for Intelligent Transportation Systems (ITS). While deep learning models, such as graph neural networks and attention-based models, have significantly advanced spatiotemporal learning, their evaluation still heavily relies on small, fixed, and clean datasets. To address this gap, we propose a data-centric framework that enhances the efficiency, robustness, and generalization of existing models through representative data selection, noise correction, and out-of-distribution (OOD)-aware data augmentation.
About Confirmation Seminars
The purpose of the confirmation milestone is to ensure that the candidate receives appropriate feedback in relation to the viability and progress of the thesis project and that the resources required to complete the program of research within the recommended timeframe are available. It is also an opportunity to ensure that the candidate has demonstrated the capacity and capability to successfully complete the thesis in a timely manner.
