Predictive analytics for supply chain resilience in urban infrastructure networks using graph convolutional networks

Area of Science:

• Urban planning and infrastructure resilience
• Complex network analysis and predictive modeling
• Supply chain management and disruption analysis

Background:

• Urban infrastructure networks are critical for supply chains but vulnerable to disruptions.
• Predicting these disruptions requires frameworks integrating network topology and flow dynamics.

Purpose of the Study:

• To propose and validate a three-stage framework for predicting urban infrastructure disruptions.
• To assess the resilience of urban networks under various failure scenarios.
• To identify key factors influencing disruption prediction and urban resilience.

Main Methods:

• Complex network analysis to map infrastructure topology.
• Graph Convolutional Network (GCN) for predictive modeling of disruptions.
• Resilience stress testing using simulated targeted and random failures.

Main Results:

• The GCN model achieved high accuracy (RMSE 5.8, R2 0.92), outperforming baseline models.
• Graph structure significantly reduced prediction error by 29.3%.
• Targeted attacks cause initial fragmentation, while random failures induce greater cumulative damage beyond 30% node removal.

Conclusions:

• The integrated framework effectively bridges flow prediction and resilience assessment.
• Temporal factors and historical flow patterns are dominant predictors of disruptions.
• Identified dual-risk nodes enable prioritized infrastructure protection for enhanced urban resilience.