Calculation and prediction of particle ratios produced from relativistic heavy-ion collisions using hadron resonance gas (HRG) and artificial neural networks (ANNs) models

Abstract This study offers a detailed investigation of several particle yield ratios generated in nucleus–nucleus collisions over a broad range of beam energies accessible at AGS, SPS, and RHIC. Two distinct frameworks are employed for this purpose: the Hadron Resonance Gas (HRG) model and Artificial Neural Networks (ANN). Particle ratios serve as essential indicators for probing the QCD medium and identifying signatures of de-confinement, especially within the energy regime examined by the RHIC Beam Energy Scan (BES) and projected for future experiments at NICA and FAIR. The observables analyzed encompass antiparticle-to-particle ratios ( K − / K + , π − / π + , p ¯ / p , Λ ¯ / Λ , Σ ¯ / Σ , Ω ¯ / Ω ), baryon-to-meson ratios ( p ¯ / π − , p / π + , Λ/ π − , Ω/ π − ), and strange-to-non-strange meson ratios ( K + / π + , K − / π − ). An ANN is trained on experimental data to learn the energy dependence of these ratios and to forecast values at unexplored collision energies. The comparative results of both the ANN and HRG approaches show strong consistency with available measurements, reinforcing the viability of ANN as a predictive modeling tool in the study of particle production, especially for future experiments.