Classification of Tokaj Wines by Ultraviolet–Visible Spectroscopy

Abstract

The suitability of UV–Vis spectrometry was evaluated for the classification of undiluted and diluted Slovak Tokaj wine samples according to style (essences, Tokaj selection, varietal and other wines), grade (quantity of cibebas), and variety (Furmint, Lipovina, and Muškát žltý) using principal component analysis (PCA), variable selection (VS), linear discriminant analysis (LDA), general discriminant analysis (GDA), and support vector machine (SVM). The individual groups of Tokaj wines differed in their production process, the quantity of cibebas used in their production, and the accorded Protected Designation of Origin status, all of which determined their price. In general, it was found that better classification was obtained based on the UV–Vis spectra of the diluted samples, VS was a more suitable algorithm for reducing the number of variables than PCA, and finally, LDA/GDA was preferred over SVM. The best total correct classification (100%) was obtained using diluted wines and VS-GDA method. To achieve this result, 45, 31, and 10 variables were needed for classification by style, grade, and variety, respectively. Thus, UV–Vis spectrometry combined with chemometrics can be widely exploited for quality control and authentication of Tokaj wines because of a relative simplicity, short-time analysis, and without considerable financial expenses.