Data fusion-driven intelligent strategy for origin traceability and quality assessment of ginger: Integrating FT-NIR, LC-MS and machine learning

The quality of ginger ( Zingiber officinale Roscoe) varies with geographical origin. This study presents an intelligent strategy for tracing ginger origin and assessing its quality by integrating liquid chromatography-mass spectrometry and Fourier transform near-infrared with machine learning. Results showed that the contents of 10 components in ginger varied among different origins. The contents of gingerols were mostly greater than 1.00 mg/g. Among these, 6-Gingerol was the most abundant compound (9.53 ± 1.24–17.21 ± 2.65 mg/g in six regions). Correlation analysis indicated that these compounds were potentially modulated by environmental factors. Principal component analysis enabled preliminary discrimination in ginger regions, while data fusion combined with k-nearest neighbor and support vector machine achieved 100% classification accuracy. Furthermore, a common bands strategy was proposed to quantify 10 components by screening shared bands for homologous compounds via wavelength optimization. Model performance was comparable to full-spectrum analysis, with the coefficient of determination for prediction around 0.8000 and the residual predictive deviation values exceeding 2.0. Overall, this strategy provides an efficient framework for ginger origin traceability and quality assessment.