Priyanka Cheema ; Racha Cheaib ; Bruce Yabsley ; Chia-Ling Hsu ; Savino Longo

21 April 2023

Abstract: We present an overview of the methodology used to develop FastBDT classifiers that suppress beam background clusters and fake photons, with the aim of using these classifiers to improve the signal-background separation power of the residual calorimeter energy E_ECL. The feature selection, hyperparameter tuning, and training of the classifiers is performed using photons from simulated BBbar events. The application of these classifiers to the B∗ → D∗lν decay is shown, with E_ECL distributions before and after classifier cuts also included. These E_ECL distributions also provide a comparison between simulated events and a subset of Belle II data amounting to 25.4 fb−1 of integrated luminosity.