stacking-ensembles

This paper presents a hybrid architecture combining FT-Transformer with gradient-boosted trees via calibration-aware stacking for customer churn prediction on structured tabular data, achieving improved F1 and AUC-ROC on a public bank churn dataset.