Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Urban Traffic Forecasting with Integrated Travel Time and Data Availability in a Conformal Graph Neural Network Framework

Mayur Patil, Qadeer Ahmed, Shawn Midlam-Mohler

TL;DR

The paper tackles urban traffic forecasting by integrating inter-station travel times into a graph neural network adjacency matrix and weighting nodes by data availability, while quantifying uncertainty with Adaptive Conformal Prediction. The proposed GCN-LSTM-ACP framework captures spatial dependencies, long-term temporal patterns, and delivers dynamic uncertainty intervals, validated through SUMO-based Monte-Carlo simulations against INRIX data. Empirically, it achieves substantial improvements over baselines (approximately $24\%$ lower MAE and $8\%$ lower RMSE) and produces prediction intervals with about 90% coverage, closely aligning with observed travel-time percentiles. This approach offers a scalable, physics-informed, and uncertainty-aware tool for real-time urban traffic management and planning, with potential applicability across diverse city contexts and data environments.

Abstract

Traffic flow prediction is a big challenge for transportation authorities as it helps plan and develop better infrastructure. State-of-the-art models often struggle to consider the data in the best way possible, as well as intrinsic uncertainties and the actual physics of the traffic. In this study, we propose a novel framework to incorporate travel times between stations into a weighted adjacency matrix of a Graph Neural Network (GNN) architecture with information from traffic stations based on their data availability. To handle uncertainty, we utilized the Adaptive Conformal Prediction (ACP) method that adjusts prediction intervals based on real-time validation residuals. To validate our results, we model a microscopic traffic scenario and perform a Monte-Carlo simulation to get a travel time distribution for a Vehicle Under Test (VUT), and this distribution is compared against the real-world data. Experiments show that the proposed model outperformed the next-best model by approximately 24% in MAE and 8% in RMSE and validation showed the simulated travel time closely matches the 95th percentile of the observed travel time value.

Urban Traffic Forecasting with Integrated Travel Time and Data Availability in a Conformal Graph Neural Network Framework

TL;DR

The paper tackles urban traffic forecasting by integrating inter-station travel times into a graph neural network adjacency matrix and weighting nodes by data availability, while quantifying uncertainty with Adaptive Conformal Prediction. The proposed GCN-LSTM-ACP framework captures spatial dependencies, long-term temporal patterns, and delivers dynamic uncertainty intervals, validated through SUMO-based Monte-Carlo simulations against INRIX data. Empirically, it achieves substantial improvements over baselines (approximately lower MAE and lower RMSE) and produces prediction intervals with about 90% coverage, closely aligning with observed travel-time percentiles. This approach offers a scalable, physics-informed, and uncertainty-aware tool for real-time urban traffic management and planning, with potential applicability across diverse city contexts and data environments.

Abstract

Traffic flow prediction is a big challenge for transportation authorities as it helps plan and develop better infrastructure. State-of-the-art models often struggle to consider the data in the best way possible, as well as intrinsic uncertainties and the actual physics of the traffic. In this study, we propose a novel framework to incorporate travel times between stations into a weighted adjacency matrix of a Graph Neural Network (GNN) architecture with information from traffic stations based on their data availability. To handle uncertainty, we utilized the Adaptive Conformal Prediction (ACP) method that adjusts prediction intervals based on real-time validation residuals. To validate our results, we model a microscopic traffic scenario and perform a Monte-Carlo simulation to get a travel time distribution for a Vehicle Under Test (VUT), and this distribution is compared against the real-world data. Experiments show that the proposed model outperformed the next-best model by approximately 24% in MAE and 8% in RMSE and validation showed the simulated travel time closely matches the 95th percentile of the observed travel time value.
Paper Structure (17 sections, 22 equations, 5 figures, 1 table, 1 algorithm)

This paper contains 17 sections, 22 equations, 5 figures, 1 table, 1 algorithm.

Table of Contents

  1. INTRODUCTION
  2. Related Work
  3. Spatio-Temporal Models
  4. Physics-Based Deep Learning Models
  5. Uncertainty Quantification
  6. Methodology
  7. Graph Network
  8. Graph Convolution Network (GCN)
  9. LSTM with Attention Mechanism
  10. Adaptive Conformal Prediction (ACP)
  11. Optimization
  12. Experiment
  13. Dataset
  14. Evaluation Metric and Baselines
  15. Results
  16. ...and 2 more sections

Figures (5)

  • Figure 1: Overall Model Architecture
  • Figure 2: Sampled Route (left); Heat Map of Adjacency Matrix (right)
  • Figure 3: Loss versus Epochs (left); MAE versus Epochs (right)
  • Figure 4: Prediction with Uncertainty Bounds
  • Figure 5: PDF of VUT travel times from Monte-Carlo SUMO simulations (left); INRIX Probe Data (right) during 7:30AM - 8:30AM