Identification of green tea varieties and origins based on fused surface enhanced Raman spectra from different extraction solvent.

This study addresses the issues of variety adulteration and origin fraud in the green tea market by establishing classification models based on fused surface-enhanced Raman scattering (SERS) spectra from three different extraction solvents. Seven green tea varieties were extracted using three solvents including water, ethanol, and hydrochloric acid to obtain single SERS spectra, and then the single SERS spectra were combined into fused SERS spectra in order to improve the discrimination capability. Various dimensionality reduction techniques for SERS spectral data such as competitive adaptive reweighted sampling (CARS), principal component analysis (PCA), t-distributed stochastic neighborhood embedding (t-SNE), and uniform manifold approximation and projection (UMAP) were applied to reduce the dimension of the fused SERS data. Finally, support vector machine (SVM), K-nearest neighbors (KNN), and random forest (RF) were used to develop classification models. The results demonstrate that fused SERS data can markedly elevate model classification accuracy relative to single SERS spectra, and the UMAP-SVM strategy achieved best performance with an accuracy rate of 98.21 % for the seven green tea varieties. Moreover, the accuracy rates for identifying the origins of Biluochun tea and Longjing tea reached 98.61 % and 97.22 %, respectively. This study establishes a simple, green, and rapid method for identifying green tea varieties and their origins based on fused SERS spectral data obtained under different extraction conditions, thereby providing a new research approach for green tea quality monitoring.