The combined application of principal component analysis and decision tree in nuclear pulse shape discrimination

Currently, classic methods and some machine learning algorithms are popular with researchers in nuclear pulse shape discrimination (PSD). Although decision tree (DT) is widely used as a classification algorithm, it is rarely found in PSD studies. In this paper, compared with the support vector machine (SVM), the effect and efficiency of the combined application of principal component analysis (PCA) and DT in PSD are reported. The results show that the DT is an efficient, effective method when combined with PCA because the accuracies are all above 90.7%, even though the difference among the pulse shapes cannot be discriminated with the naked eye. The testing speeds of the DT with PCA are over 36 times those of SVM, and the processing speeds are sufficiently high for most situations.