Using machine learning to predict gamma shielding properties: a comparative study

Abstract

This study employed machine learning (ML) algorithms to predict the linear attenuation coefficients (LACs) of materials in inorganic scintillation detectors, which are crucial for evaluating self-shielding properties. Predictions from various ML models were compared with results from the Phy-X/PSD program across different photon energies. The Gradient Boosting Regressor (GBR) model was identified as the most accurate model, achieving a testing set accuracy of 96.40%. This research showcases the potential of ML for efficiently and accurately estimating LACs, with the GBR model showing promise for applications in radiation detection and material science.

Keywords

• linear attenuation coefficients
• inorganic scintillation detectors
• Phy-X/PSD
• machine learning
• gradient boosting regressor
• PyCaret