Imbalanced classification of antineutrino and cosmic muon in segmented plastic scintillator antineutrino detector

Inverse beta decay (IBD) in plastic scintillators is one of the most commonly used methods for detecting reactor antineutrinos. Cosmic muon signals due to the IBD compared to those generated by antineutrinos are still the main challenge in these types of detectors. The IRAND (IRan ANtineutrino Detector) is currently being designed and implemented with the constraint of reducing the required hardware, and at the same time, improving the antineutrino detection efficiency. Imbalanced classification is one of the software methods in machine learning that deals with imbalanced data, such as muon and antineutrino. Using the IRAND-Sim simulation package based on the Geant۴ toolkit presented in our previous research, the spectra and angular distribution of antineutrinos and muons can be calculated. However, in this study, the memory management techniques to handle the dataset due to a large number of muons have been used, and also two separate methods have been used in the imbalanced classification for discriminating muon and antineutrino events. The results show that this approach by combining real and simulated data is very efficient, and the imbalanced nature can be reduced to achieve better classifier performance.

کلیدواژه ها:

Segmented plastic scintillator antineutrino detector ، Machine Learning ، Geant۴ toolkit ، Muon reduction ، Imbalanced classification