Mostafi, Sifatul (Ontario Tech University), Elgazzar, Khalid (Ontario Tech University), Abdellatif, Tamer Mohamed (Canadian University Dubai, UAE)

ForeCAST-GAT: Context-Aware Spatio-Temporal Graph Attention Transformer for Event-Driven Traffic Forecasting

Scheduled for presentation during the Video Session "On-Demand Video Presentations" (VP-VP), Saturday, November 22, 2025, 08:00−18:00, On-Demand Platform

2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), November 18-21, 2025, Gold Coast, Australia

This information is tentative and subject to change. Compiled on April 2, 2026

Abstract

Accurate traffic forecasting is critical for intelligent transportation systems to effectively manage disruptions on highways caused by accidents, lane closures, and severe weather events. While traditional traffic forecasting models primarily rely on historical traffic data, they often fail to account for the sudden impacts of such events. To address this gap, we introduce ForeCAST-GAT, a Context-Aware Spatio-Temporal Graph Attention transformer for event-driven traffic forecasting. By embedding both spatio-temporal dependencies and event-specific contextual features into each highway segment’s node representation, ForeCAST-GAT dynamically adapts to changing traffic patterns during events. The model leverages Graph Attention Networks (GATs) to capture spatial relationships among highway segments and utilizes transformer-based mechanisms to model temporal dynamics, enabling it to account for both inherent traffic speed patterns and external disruptions. This integration improves the model's forecasting accuracy during event-driven scenarios. Evaluations conducted on real-world traffic and event data collected from Highway 401 reveal that ForeCAST-GAT achieves up to a 35.9% reduction in MAE and a 20.2% reduction in RMSE compared to existing models that do not consider event-specific features.