Abstract
Identifying the geographic origin of a wine is of great importance, as origin fakery is commonplace in the wine industry. This study analyzed the mineral elements, volatile components, and metabolites in wine using inductively coupled plasma-mass spectrometry, headspace solid phase microextraction gas chromatography-mass spectrometry, and ultra-high-performance liquid chromatography-quadrupole-exactive orbitrap mass spectrometry. The most critical variables (5 mineral elements, 13 volatile components, and 51 metabolites) for wine origin classification were selected via principal component analysis and orthogonal partial least squares discriminant analysis. Subsequently, three algorithms-K-nearest neighbors, support vector machine, and random forest -were used to model single and fused datasets for origin identification. These results indicated that fused datasets, based on feature variables (mineral elements, volatile components, and metabolites), achieved the best performance, with predictive rates of 100% for all three algorithms. This study demonstrates the effectiveness of a multi-source data fusion strategy for authenticity identification of Chinese wine.