Spectroscopic techniques combined with DD-SIMCA model and explainable artificial intelligence for rapidly and accurately identifying the quality of Astragali Radix

Astragali Radix (AR) is a traditional Chinese medicine (TCM) widely used worldwide for its nutritional and medicinal benefits, but it is facing increasing quality issues. This article developed a novel method for quickly and accurately identifying AR’s quality. In this study, spectroscopic techniques combined with data-driven soft independent modelling of class analogy (DD-SIMCA) and explainable artificial intelligence (XAI) were used to determine AR’s geographic origins and predict its antioxidant activity. The results showed that preprocessed near-infrared (NIR) or raw visible (VIS) spectra with DD-SIMCA could accurately identify AR’s authentic regions, with 100 % sensitivity, specificity, and accuracy. Additionally, XAI identified 80 features from NIR spectra and 33 from VIS spectra, both strongly correlated with AR’s antioxidant activity. Fusing these features and integrating them with support vector machine led to significantly better model performance, with R p 2 of 0.9760, RPDP of 6.7447, RMSEP of 1.3824, and MAEP of 1.1088. Overall, this study provided valuable insights for the quality assessment of TCMs and other medicinal plants.