Predicting (n,3n) nuclear reaction cross-sections using XGBoost and Leave-One-Out Cross-Validation

Area of Science:

• Nuclear Physics
• Computational Science
• Materials Science

Background:

• Accurate prediction of nuclear reaction cross-sections is vital for nuclear medicine, energy, and materials science.
• Existing theoretical and empirical methods have limitations in predicting these cross-sections.

Purpose of the Study:

• To develop a robust machine learning model for predicting (n,3n) nuclear reaction cross-sections.
• To leverage the XGBoost algorithm for enhanced prediction accuracy.

Main Methods:

• Utilized a comprehensive dataset of experimental cross-sectional values.
• Employed Leave-One-Out Cross-Validation (LOOCV) for feature selection and hyperparameter optimization.
• Prepared data by normalizing values and handling missing data for robust model training.

Main Results:

• The XGBoost model demonstrated high accuracy in predicting (n,3n) nuclear reaction cross-sections.
• Evaluated model performance using Mean Squared Error (MSE), Mean Absolute Error (MAE), and reduced Chi-Square.
• Compared XGBoost predictions with TALYS 1.95, TENDL, and phenomenological models, showing improved accuracy.

Conclusions:

• Machine learning, specifically XGBoost, offers a powerful approach to overcome limitations in nuclear data analysis.
• The developed model provides accurate predictions, advancing the field of nuclear reaction cross-section analysis.