Model Interpretation and Data-centric Modeling for Advanced traffic Prediction (MINDMAP)
About
Traffic prediction is a crucial research area for monitoring and managing the dynamic demand and supply patterns. While previous studies have demonstrated the effectiveness of deep learning, particularly Graph Neural Networks (GNNs), in achieving accurate traffic forecasts, much of the focus has been on optimizing model accuracy using standardized datasets. Transfer learning presents an opportunity to adapt pre-trained traffic prediction models to new datasets or study areas. However, real-world data often deviate from ideal conditions, posing challenges for model transfer in practical applications. Additionally, the interpretability of GNN-based traffic prediction models remains largely unexplored.
The MINDMAP project aims to develop and evaluate both theoretical and practical methods for data-efficient and interpretable transfer of traffic prediction models. A data-centric approach will ensure systematic transformation of raw data into meaningful representations, enhancing the adaptability of pre-trained models. Simultaneously, interpretability techniques will offer transparency into the model’s decision-making process, identifying transferable and non-transferable elements of demand and supply.
By improving transfer learning and interpretability in traffic prediction, MINDMAP seeks to enhance existing research and support the practical application of these methods in real-world scenarios.
Partners
The chair of Transportation Systems Engineering (TSE) of TUM is collaborating on this international project with the group of Prof. Dr. Jiwon Kim from the University of Queensland, Australia, and the Data Analytics and Machine Learning (DAML) group of TUM Prof. Dr. Stephan Günnemann. The research team will approach the ambitious research goals set out in this proposal from the perspectives of transferability, interpretability, and data-centric perspectives, using distinct but complementary methodologies and access to different sets of collaborators and resources while striving for a common goal.
The doctoral researcher from at TSE is funded for this project by the International Graduate School of Science and Engineering (IGSSE).
Contact
Prof. Dr. Constantinos Antoniou, MSc. Soban Lone (PhD Candidate)