A0411
Title: A conformal prediction framework for network systems
Authors: Rui Luo - City University of Hong Kong (Hong Kong) [presenting]
Abstract: Urban monitoring and sensor networks require reliable uncertainty quantification, yet methods with formal statistical guarantees are scarce. This is addressed by integrating graph neural networks (GNNs) with conformal prediction to create a framework for trustworthy, uncertainty-aware network analysis. This report details the advances in three core applications: conformal node classification for anomaly detection, conformal edge regression for traffic forecasting, and conformal route planning for risk-sensitive navigation. The methods are validated on 15 real-world datasets, demonstrating robust performance across diverse network tasks. Finally, future applications are discussed in critical urban systems, including cybersecurity and intelligent transportation.
