Machine Learning in High-Momentum Particle Identification in the MPD Experiment

One of the main challenges in particle identification (PID) in the MPD experiment at the NICA accelerator complex is classification of particle species in high momentum range where conventional methods, such as n-sigma lose efficiency. This study is devoted to application of gradient boosted decision trees (GBDT) for identification of six particle types which were produced in bismuth–bismuth simulated collisions at $$\sqrt {{{s}_{{NN}}}} = 9.2$$ GeV. The XGBoost algorithm was compared to the $$n$$ -sigma and blind methods, evaluating efficiency and contamination. Results show that XGBoost provides significant PID performance in momentum ranges where feature overlap limits traditional techniques, highlighting the potential of machine learning to improve MPD analyses.