Enhanced cross-category models for predicting the total polyphenols, caffeine and free amino acids contents in Chinese tea using NIR spectroscopy

Abstract

Total polyphenols (TP), caffeine, and free amino acids (FAA) constitute tea taste. The feasibility of developing a cross-category model for predicting TP, caffeine and FAA contents in Chinese black, dark, oolong, and green teas, was investigated. Diffuse reflectance spectra (4000–10,000 cm−1) of tea were collected using Fourier transform near-infrared (NIR) spectroscopy, and a hybrid method was applied to enhance characteristic signals. Random frog and competitive adaptive reweighted sampling (CARS) were used to select key variables for partial least squares (PLS) calculation. For calibration, the best predictive performance was achieved by the CARS-PLS models. The coefficients of determination and root mean squared errors in the prediction set were 0.994 and 0.595 for TP, 0.986 and 0.070 for caffeine, and 0.993 and 0.063 for FAA, respectively. The results highlight the potential of NIR coupled with chemometrics for the simultaneous testing of TP, caffeine and FAA contents in Chinese tea from different categories.