ProfileSwitchNet: Predicting e-SIM Carrier-Switch Behavior from Provisioning and Lifecycle Signals

Abstract

This work introduces ProfileSwitchNet, a predictive model for forecasting e-SIM carrier-switch behavior based on historical provisioning and lifecycle data. Each e-SIM profile is represented by features such as device type, initial and subsequent carrier profiles, activation counts, roaming flags, region, data plan class, and recorded status changes over time. A sequence-aware model (e.g., temporal gradient boosting or a recurrent neural network) is trained to estimate the probability that a profile will be deactivated or switched to a different carrier within a future time window. The framework evaluates performance using ROC-AUC, precision–recall, and calibration metrics, and analyzes feature importance to identify the strongest drivers of switching, such as device category or activation failure bursts. ProfileSwitchNet is intended to help operators design better retention offers, optimize roaming strategies, and understand early signals of churn in eSIM ecosystems.