Objective <p>This paper aims to propose an intelligent discrimination method for distinguishing between n and γ pulses, based on multi-featureparameters and the K-nearest neighbor-Linear discriminant analysis (KNN-LDA) algorithm.</p> Methods <p>Multi-feature parameters, incorporating both time and frequency domains, are established using the pulse shape discrimination(PSD) principle. An automatic feature extraction system is designed to intelligently extract the optimal distributions of multi-featuresfrom pulse data. A feature criterion is then constructed based on these multi-feature distributions. The reliable data are used to trainthe KNN-LDA model with regression optimization and dimensionality reduction. The remaining data are used as a test set forclassification.</p> Results <p>The experimental results show that the Figure of Merit (FOM) value of the KNN-LDA model reaches 3.07 in the high-energy domain (40 keV), which is 245% higher than that of the charge comparison method (CCM). In the low-energy domain (≤ 40 keV), the FOMvalue is 2.64, which is 355% higher than that of the CCM.</p> Conclusions <p>The proposed method demonstrates excellent n/γ discrimination capabilities, especially in the low-energy domain. This provides anew approach for solving the challenging problem of low-energy domain discrimination using supervised learning models</p>