Scalable machine learning framework for predicting critical links in urban networks - Article Abstract

Aug 4, 2026

Efficient identification of critical links in urban road networks underpins traffic management, infrastructure planning, and resource allocation. Conventional simulation-based and robustness-index methods become computationally prohibitive at large scale and often rely on limited structural or functional signals. This study presents a scalable machine learning framework that trains on only 20% of network links and predicts criticality for the rest with about 7% mean percentage error, combining structural, functional, and new dynamic traffic features. Tested on LuST and MoST datasets, the framework achieves strong single-city precision and robust cross-city generalization, with Random Forest and Gradient Boosting leading. The results support scalable evaluation with limited data and suggest future directions including domain adaptation and temporal modeling.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages13

Reading Time33 min

AccessFree to download

Introduction

Limitations of existing robustness and simulation methods

Limits of structural-only or functional-only predictive models

Proposed scalable machine learning framework and novel indices

Journal of Innovation & Knowledge 10 (2025) 100715 | Scalable machine learning framework for predicting critical links in urban networks
Nourhan Bachira ,∗, Chamseddine Zakib,∗, Hassan Harbb, Roland Billenaa GeoScITY, Spheres Research Unit, University of Liège, 4000 Liège, Belgium
b College of Engineering and Technology, American University of the Middle East, Egaila 54200, Kuwait | | | | | --- | --- | --- | --- | | A R T I C L E I N F O | | A B S T R A C T | | | JEL classification:
C51 C52 C53 L91 R42
Keywords:
Link criticality Urban traffic networks Machine learning Traffic management Random Forest Gradient Boosting | | Efficient identification of critical links in urban road networks is essential for optimizing traffic management, infrastructure planning, and resource allocation. Existing methods, such as simulation-based approaches, are computationally expensive and often impractical for large-scale networks. This study proposes a scalable machine learning framework capable of training on a subset of network links (20%) and predicting the criticality of remaining links with approximately 7% percentage mean error. The framework integrates structural, functional, and newly proposed features, offering a comprehensive representation of road network dynamics. Validated on two diverse datasets, namely, Luxembourg (LuST) and Monaco (MoST), the framework achieves high precision (∼72% and ∼73% in single-city scenarios) and robust cross-city performance (∼70% for LuST → MoST and ∼66% for MoST → LuST). Random Forest and Gradient Boosting emerged as the topperforming models, consistently delivering the best precisions and lowest number of errors. The inclusion of dynamic traffic metrics and advanced preprocessing techniques further enhanced predictive accuracy and generalization capabilities. This study highlights the potential of machine learning for scalable critical link evaluation, demonstrating its applicability to large-scale networks with limited data. The findings provide actionable insights for urban traffic management and open pathways for future research, including domain adaptation, temporal modeling, and integration with real-time systems. | | 1. Introduction The evaluation of critical links within urban road networks is a pivotal task for transportation planning and traffic management. Accurate identification of these links is essential for ensuring efficient traffic flow, prioritizing infrastructure maintenance, and optimizing resource allocation (Shapouri, Fuller, Wolshon, & Herrera, 2023). Traffic networks are increasingly complex, requiring innovative solutions to maintain their functionality, particularly as urban populations and vehicle usage continue to rise. Effective criticality analysis ensures cities can mitigate congestion, prioritize repairs, and enhance safety, thereby supporting both economic growth and quality of life for urban dwellers. Traditional methods, such as those based on the network robustness index (NRI), assess the criticality of links by analyzing the impact of removing individual links on key traffic metrics such as total travel time and vehicle throughput. However, although effective in smallscale scenarios, these approaches are computationally prohibitive for large-scale networks. For example, evaluating networks with thousands of edges can require several hours of simulation per link, making realworld application infeasible. These methods often rely on simulations that do not scale well with increasing network size or complexity, leading to delays in actionable insights. Moreover, existing methods described as predictive models in the literature often rely exclusively on structural or functional features of road networks, such as road type, length, or traffic density. This dichotomy fails to capture the interplay between static road attributes and dynamic traffic behaviors, leading to incomplete assessment of link criticality. Notably, no existing research

FAQs

Why is predicting critical links important in urban networks?

It enables efficient traffic flow, prioritizes infrastructure maintenance, and improves resource allocation, supporting congestion mitigation and safety enhancement.

What makes the proposed framework scalable compared with traditional methods?

It trains on only 20% of network links and predicts criticality for the remaining links, avoiding exhaustive per-link simulations that become impractical for large networks.

Which models performed best in the study?

Random Forest and Gradient Boosting emerged as the top-performing models, delivering the best precision and the lowest number of errors across evaluations.

Aug 4, 2026

Upload by River Wang