The accurate determination of the mass composition and energy spectrum of primary cosmic rays (PCRs) around the knee ( \(\sim 3\,\textrm{PeV}\) ) is vital for understanding their origin, acceleration mechanisms, and interactions with the interstellar medium. The GRAPES-3 experiment in Ooty, India, utilizes a dense array of scintillator detectors and a large-area Muon Telescope to study PCRs from several \(\textrm{TeV}\) to \(\sim 10 \, \textrm{PeV}\) . Using Boosted Decision Trees (BDTs), we study classification of light primaries (protons and helium) from heavier nuclei, with a dedicated classifier to further separate protons from helium. The study is performed on Monte-Carlo simulated data which will later be applied to GRAPES-3 observations to extract primary compositions and energy spectra, providing insights into cosmic ray origins and propagation around the knee region.