Machine-Learning-Assisted Analysis of Dimuon Events using CMS Open Data

Proton–proton collisions at 8 TeV. This work analyzes event-level CMS Open Data, pub- licly available via CERN. The focus remains squarely on the dimuon invariant-mass distribution. We tested whether a standard Random Forest classifier could effectively isolate signal from background using only eight kinematic features derived from recon- structed muons. Signal/background labeling employed a simple mass window for the dimuon system. Model performance assessment involved ROC curves, observed score dis- tributions, and feature importance studies. Predictably, muon transverse momenta drove the classification decision almost entirely; angular geometry provided secondary, yet help- ful context. This is not a formal CMS measurement. Rather, it constitutes a reproducible baseline demonstrating how machine learning techniques apply to open physics datasets.