Abstract: Traffic congestion in urban and regional transportation networks represents a critical challenge affecting economic productivity, environmental sustainability, and quality of life. Recent advances in spatiotemporal deep learning have enabled sophisticated predictive models capable of capturing complex dependencies in traffic data across both spatial and temporal dimensions. This paper provides a comprehensive examination of spatiotemporal deep learning methodologies applied to predictive congestion management, synthesizing recent developments in graph neural networks, recurrent architectures, attention mechanisms, and hybrid frameworks. The analysis encompasses architectural innovations including spatiotemporal graph convolutional networks, graph attention networks, temporal point processes, and physics-guided neural networks. Empirical evidence from benchmark datasets demonstrates that these approaches achieve substantial improvements in prediction accuracy compared to traditional methods, with mean absolute percentage errors frequently below 5% for short-term forecasting horizons. However, significant challenges persist, including performance degradation during congested periods, computational complexity for large-scale networks, and limited model interpretability. The synthesis reveals that hybrid architectures combining multiple spatiotemporal modeling paradigms consistently outperform single-method approaches. Future research directions include enhanced interpretability mechanisms, real-time deployment optimization, integration of multimodal transportation data, and development of adaptive models capable of maintaining accuracy during extreme congestion events.
Keywords: spatiotemporal deep learning, traffic congestion prediction, graph neural networks.
Title: Spatiotemporal Deep Learning for Predictive Congestion Management in Urban and Regional Transportation Networks
Author: Oghenerunor Angel Ederewhevbe, Ejemen Obozokhae
International Journal of Civil and Structural Engineering Research
ISSN 2348-7607 (Online)
Vol. 14, Issue 1, April 2026 - September 2026
Page No: 35-45
Research Publish Journals
Website: www.researchpublish.com
Published Date: 03-June-2026
