Identification of Black Tea from Four Countries by Using Near-infrared Spectroscopy and Support Vector Data Description Pattern Recognition

Abstract

In this study, black teas from four different countries were successfully identified using near-infrared (NIR) spectroscopy combined with the Support Vector Data Description (SVDD) algorithm. The original spectra of tea ranged in wavelength from 12500 to 4000 cm^-1. We used SVDD to optimize the parameters and calibrate the discrimination model. As a comparison, the K-Nearest Neighbor algorithm (KNN) and Partial Least Square (PLS) were also used in this study. Compared with the KNN and PLS classifications, the SVDD model was better able to deal with imbalance training samples and outperformed the other models in the prediction set. The optimal SVDD model was achieved with principal components (<em>PC</em>) = 5. Identification rates were 96.25% in the training set and 92.50% in the prediction set. These results indicate that NIR spectroscopy combined with SVDD is a useful tool in building a one-class calibration model for discrimination of black tea from different countries.