Classifying Muon Neutrino Interactions in the MINERνA Detector using Gradient-Boosted Decision Trees

Gradient-boosted decision trees are used to classify muon neutrino interactions in the MINERνA detector, with the goal of identifying charged-current quasielastic-like scattering events on hyrdocarbon. Decision trees are a popular method used in machine learning for the purposes of both classification and regression, and involve the partitioning of a feature space into rectangular regions for which discrete or continuous values are assigned. Boosting assists this method by forming an ensemble of decision trees that collectively outperforms any individual tree. Packages being used to implement gradient-boosted decision trees are provided by the Toolkit for Multivariate Data Analysis (TMVA) within the ROOT data analysis framework. Preliminary results demonstrate both higher efficiencies and higher purities compared to those produced by conventional cut-based selection methods.