Classification of Juniper-Flavoured Spirit Drinks by Multivariate Analysis of Spectroscopic and Chromatographic Data

Abstract

This study compares the use of UV absorption, excitation–emission matrix (EEM) fluorescence and synchronous fluorescence spectroscopy and HPLC with fluorescence detection combined with principal component analysis (PCA), parallel factor analysis (PARAFAC) and linear discriminant analysis (LDA) for distinguishing between commercial samples of Slovak, Belgian, German, Czech and British juniper-flavoured spirit drinks. Overall, 97 %, 88 % and 79 % of samples were properly classified by applying the LDA to the first five principal components of the PCA performed on the synchronous fluorescence spectra (constant wavelength difference 10 nm, 250–450 nm), UV absorption spectra (250–325 nm) and the areas of eight common HPLC peaks, respectively. EEM fluorescence spectroscopy could not discriminate British drinks, while HPLC failed to discriminate Belgian samples. When the areas of eight common HPLC peaks were used as model parameters instead of the five PCs initially used in LDA, the method accuracy was enhanced significantly and 97 % of predicting samples were properly classified.