Multi-components of green tea from different origins were identified by UPLC–Orbitrap–MS/MS, including alkaloids, amino acids, catechins, flavones, flavone glycosides, phenolic acids and theaflavins. Quantitative chemical profiles of 72 samples of Yunnan green tea (YGT) and Hunan green tea (HGT) established the influence of origin on green tea. In addition, three variable selection methods, based on partial least squares-discriminant analysis (PLS-DA), were employed to screen characteristic components of YGT and HGT. The results of variable importance on projection (VIP) method showed better performance than coefficients β and the least absolute shrinkage and selection operator (LASSO) for this dataset. Three characteristic components, named, gallocatechin gallate (GCG), epicatechin-(4β → 8)-epigallocatechin-3-O-gallate, and vitexin were selected by VIP, with a correct rate of 98.61% and an AUC value of 0.9706. The results indicated that the combination of UPLC–Orbitrap–MS/MS and chemometrics could serve as a valid strategy to analyze complex analytical objects, such as green tea.