Skip to main content
ICTD 2024 Main banner
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.

Abstract for presentation (Poster or Podium)

AI in Transportation

Quality Control of Traffic Data with Augmented Spatiotemporal Contexts Using Graph Neural Networks

Monday, June 17, 2024
1:30 PM - 2:45 PM ET
Room: Galleria

    Primary Author(s)

  • YH

    Yongcan Huang

    Graduate Research Assistant
    University of Georgia
    Athens, GA, United States

    • Presenter(s)

  • YH

    Yongcan Huang

    Graduate Research Assistant
    University of Georgia
    Athens, GA, United States

    • Co-Author(s)

  • JY

    Jidong J. Yang, PhD

    Associate Professor
    University of Georgia
    Athens, GA, United States

    • Submitter(s)

  • JY

    Jidong J. Yang, PhD

    Associate Professor
    University of Georgia
    Athens, GA, United States

Abstract: Quality control of traffic data plays a crucial role in supporting various engineering practices in the transportation sector. In our prior research, we showcased the effectiveness of a two-stage Graph Neural Networks (GNN) model for quality control of sequential traffic data. The model initially clusters traffic sensors based on their geospatial locations and temporal traffic patterns encoded in time sequence data. Subsequently, it identifies anomalous patterns at the cluster level by meticulously extracting spatial-temporal dynamics among these sensors. Notably, our previous methodology solely incorporated traffic count data as node content, leading to the formation of oversized clusters and the oversight of crucial traffic dynamics related to specific contexts and facility functional classifications. To enhance our sensor clustering analysis and improve our existing GNN-based quality control approach, we have expanded the node content within the traffic sensor network graph. This expansion includes the integration of contextual details and differentiation of passenger and freight flow patterns by segregating truck traffic from non-truck traffic. Our experimental findings confirm that this augmentation of node content results in enhanced clustering outcomes, consequently bolstering the reliability of the data quality control process and surpassing the performance of our previous method.

Learning Objectives:

Attendees can expect to learn the following from this session:
  • Learn how the graph neural networks can be used to model network traffic flow.
  • Understand the importance of contextual information in recognizing anomaly traffic flow patterns in the tightly coupled spatial and temporal space.
  • Learn the implementation details of graph neural networks for sensor data quality control.