A1283
Title: Normalized functional conformal prediction for adaptive anomaly detection in spaceborne ground displacement data
Authors: Teresa Bortolotti - Politecnico di Milano (Italy) [presenting]
Alessandra Menafoglio - Politecnico di Milano (Italy)
Simone Vantini - Politecnico di Milano (Italy)
Abstract: Monitoring spaceborne ground displacement in areas exposed to natural hazards requires statistical tools capable of detecting localized anomalies under complex dynamics. Leveraging functional conformal prediction, a framework for anomaly detection is proposed, providing distribution-free control of false alarms under exchangeability. The effectiveness of such methods critically depends on the tightness of the conformal prediction bands, as overly conservative bands may be blind to small but meaningful deformations. To address this, a normalized functional conformal approach is introduced that incorporates a covariate-dependent variability function, capturing heteroskedasticity and enabling locally adaptive bands across both the functional domain and different input conditions. This localized adaptivity yields prediction regions that are as informative as possible while retaining validity guarantees and enhances the sensitivity of the monitoring tool to anomalies. The effectiveness of the proposed methodology is demonstrated on ground displacement data covering the area of the Phlegraean Fields, where early and accurate detection of anomalous dynamics is crucial for mitigating risks associated with volcanic and bradyseismic activity.
