Exploration of New Analysis Techniques Using Machine Learning in Dilepton Analysis

Dilepton production in proton-proton collisions arises from a variety of processes, including the Drell-Yan process, decays of J/ψ and Y resonances, open charm and open bottom production, as well as light meson decays. The overlap of the resulting signals makes isolating and studying individual processes challenging; to address this issue, machine learning techniques can be used to enhance signal-to-background data extraction. A gradient boosting algorithm, XGBoost, was used to train a machine learning model to classify dilepton pairs according to their origin, in order to separate contributions from each process with improved precision. This method has the potential to surpass traditional techniques by producing cleaner and more representative datasets. With improved separation between processes, researchers can more effectively study the dynamics of dilepton production and the individual processes that contribute to dilepton production. Simulation results using Pythia8 will be presented in this poster presentation.