Determination of (p, n) reaction cross-section for various nuclei at 7.5 MeV by using machine learning models.

Abstract

This study investigates the prediction of (p, n) reaction cross-sections for various nuclei at 7.5 MeV using machine learning models. A dataset of 91 instances, containing key nuclear properties such as mass number (A), proton number (Z), neutron number (N), and the asymmetry term ((N-Z)/A²), was utilized. Various machine learning techniques, including Random Forest, Support Vector Regression (SVR), Gradient Boosting, Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), K-Nearest Neighbours, Multiple Linear Regression and Ensemble Model were employed. Model performances were evaluated using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and R² metrics. Among the models, ensemble methods, SVR, and boosting-based approaches demonstrated superior predictive capabilities, effectively capturing nonlinear relationships between nuclear properties and cross-sections. Results highlight the significance of the asymmetry term in enhancing prediction accuracy. This study underscores the potential of machine learning as a robust tool for nuclear physics applications, particularly in understanding and predicting nuclear reaction cross-sections.