Comparing Machine Learning Methods for Classifying Radioactive Elements Using Prompt-Gamma-Ray Neutron Activation Data

Researchers have developed machine learning algorithms that can effectively identify radioactive materials using prompt-gamma-ray neutron activation data. Prompt-gamma activation analysis (PGAA) is a non-destructive radio-analytical technique that can detect elements using gamma rays from neutron capture. However, PGAA can be hindered by long detection times and a high rate of false positives. In this study, six different machine-learning algorithms were used to classify radioactive elements based on PGAA energy spectra, including decision trees, random forest, Adaboost, support vector machine, and k-nearest neighbours. The results showed that the tree-based algorithms outperformed the others, with Adaboost being the preferred classifier for analysing PGAA spectral information due to its high recall and minimal false negatives. This study shows promise for reducing false alarm rates in detecting illicit radioactive materials in nuclear forensics, which is crucial for nuclear security to combat threats from nuclear terrorism.