Multi-element determination via optimized microwave digestion-ICP-MS, NIR-PLS quantitative prediction, and AI-based geographical origin traceability of Taiping Houkui

To support both six-origin identification and core/non-core differentiation of Taiping Houkui tea, this study established an analytical framework integrating mineral element analysis with near-infrared spectroscopy (NIR). Microwave digestion was optimized (180 °C, 20 min, HNO 3 -H 2 O 2 5:3 v/v), enabling efficient extraction of 11 mineral elements (Ca, Mg, K, Mn, Fe, Na, Cu, Zn, Cr, Pb, Cd) with recoveries of 81.2-114% and RSD <5%. The PLS models developed from the NIR spectra predicted the contents of 11 mineral elements, with test-set coefficients of determination (R 2 ) ranging from 0.9131 to 0.9685. For origin classification, Naive Bayes models using mineral element data and fused data achieved 91.7% and 83.3% six-origin accuracy (macro-averaged F1: 0.9179, 0.8135); element-based and fusion-based models both attained 94.4% core/non-core accuracy (F1: 0.9444). A 1D-CNN trained on preprocessed NIR spectral data achieved 97.2% origin accuracy (F1: 0.9720) and 94.4% core/non-core accuracy (F1: 0.9443). This work provides efficient technical support for rapid traceability and quality control of Taiping Houkui tea.