Fast, reliable Panax ginseng age classification via multispectral NIR-FTIR-RS data fusion and chemometrics

Panax ginseng is widely used in traditional medicine, but its quality varies greatly with growth age, making reliable age identification essential for ensuring authenticity and medicinal value. However, existing methods—such as morphological inspection and single-spectrum techniques—are often subjective or limited in discriminatory power, highlighting the need for a rapid and accurate alternative. In this study, we integrate Near-Infrared Spectroscopy (NIR), Fourier Transform Infrared Spectroscopy (FTIR), and Raman Spectroscopy (RS) with chemometric models, including Partial Least Squares Discriminant Analysis (PLS-DA), Support Vector Machine (SVM), and Grey Wolf Optimized SVM (GWO-SVM). Leveraging single-spectrum data and multi-source data fusion, we develop two qualitative classification models for Panax ginseng using 226 samples (aged 1–5 years) collected from the primary production areas of Northeast China. The optimal model for five-class (1–5 years) classification was the Mid-Level Data Fusion (MDLF) strategy combining NIR-FTIR-RS with SVM, achieving an F1 score of 0.87. For distinguishing non-medicinal (1–3 years) from medicinal (4–5 years) ginseng, the best model was Low-Level Data Fusion (LLDF) using FTIR-RS with GWO-SVM, achieving an F1 score of 0.99. These findings demonstrate that multi-spectral fusion combined with machine learning provides a promising, green, and efficient methodology for Panax ginseng age authentication, offering strong potential for quality assessment, though further multi-regional validation is required for broad market application. • Rapid ginseng age classification is achieved via NIR-FTIR-Raman fusion. • MDLF-SVM model effectively identifies 1–5-year-old ginseng. • LLDF-GWO-SVM reliably separates medicinal and non-medicinal ginseng. • Spectral fusion offers a promising, green approach for ginseng assessment. * LLDF: Low-Level Data Fusion, MDLF: Mid-Level Data Fusion