Publication Date
Spring 2025
Degree Type
Master's Project
Degree Name
Master of Science in Computer Science (MSCS)
Department
Computer Science
First Advisor
Katerina Potika
Second Advisor
Fabio Di Troia
Third Advisor
Ankita Mohalkar
Keywords
Variational Graph Autoencoders (VGAEs), Graph Convolutional Networks (GCNs), Convolutional Neural Networks (CNNs), Temporal Graph Representation Learning, Traffic Forecasting.
Abstract
Traffic forecasting is important for improving transportation systems by enabling better traffic management, congestion reduction, and urban planning. However, predicting traffic accurately is challenging due to the strong spatial dependencies between different road segments and the temporal changes in traffic patterns over time. Traditional time-series and graph models often struggle to capture both of these aspects effectively. In response, recent research has focused on temporal graph representation learning methods that jointly consider spatial relationships and temporal features in networks. This project proposes a hybrid model called VSET-Nets (VGAE Spatial Embedding for Temporal Networks) that employs Variational Graph Autoencoders (VGAEs) for learning spatial embeddings and Convolutional Neural Networks (CNNs) for capturing temporal features at each node. The model was evaluated using a benchmark dataset, containing traffic speed data collected from sensors in California’s District 7. Experimental results show that almost all of the node-specific regressors trained using our approach outperformed the baseline GCN model, with around 95% of nodes achieving better forecasting accuracy. These findings suggest that modular spatio-temporal modeling can offer a promising direction for building more effective traffic forecasting systems.
Recommended Citation
Dhapre, Mrunmayee, "Traffic Forecasting with VSET-Nets: A VGAE Spatial Embedding for Temporal Networks Approach" (2025). Master's Projects. 1533.
https://scholarworks.sjsu.edu/etd_projects/1533
