BDT application to dark photon searches in H->y+yd with the ATLAS detector at the LHC

This study investigates the application of a Boosted Decision Tree (BDT) model in the classification of signal events and background noise, specifically within the context of exploring dark photon phenomena. The model was optimized using hyperparameter tuning techniques, including adjustments to parameters such as num_leaves, feature_fraction, and max_depth, aimed at enhancing predictive accuracy and model performance. A thorough analysis of feature importance was performed to identify key drivers that significantly impact the model's classification capabilities. The findings demonstrate the BDT model's potential in identifying and classifying events related to dark photons, with the results providing valuable insights into the model's effectiveness in distinguishing signal events from background noise. Future efforts will focus on further refining the model and expanding the dataset to improve its classification capabilities in the ongoing investigation of dark photons.