The application of machine learning techniques to data from solid-state nuclear track detector CR-39

Abstract The study presents an application of classification-based machine learning techniques to a dataset comprising the radiation dose measurements with solid-state nuclear track detectors of poly(allyl diglycol carbonate) type. The detectors were irradiated with alpha particles and fast neutrons in various experimental configurations making the final dataset complex and suitable for machine learning methods. The most suitable experiment is chosen as a stepping stone, and the proposed evaluation method is tested. The detectors are analysed with the commercially available TASLImage system and the final performance in dose determination is compared to the machine learning efforts. Moreover, the uncertainty quantification algorithm is applied to better judge the future applicability of the new evaluation method.