Characterization of Slovenian wines using Multidimensional data analysis from simple enological descriptors

Determination of the product's origin is one of the primary requirements when certifying a wine's authenticity. Significant research has described the possibilities of predicting a wine's origin using efficient methods of wine components' analyses connected with multivariate data analysis. The main goal of this study was to examine the discrimination ability of simple enological descriptors for the classification of Slovenian red and white wine samples according to their varieties and geographical origins. Another task was to investigate the inter-relations available among descriptors such as relative density, content of total acids, non-volatile acids and volatile acids, ash, reducing sugars, sugar-free extract, SO2, ethanol, pH, and an important additional variable - the sensorial quality of the wine, using correlation analysis, principal component analysis (PCA), and cluster analysis (CLU). 739 red and white wine samples were scanned on a Wine Scan FT 120, from wave numbers 926 cm(-1) to 5012 cm(-1). The applied methods of linear discriminant analysis (LDA), general discriminant analysis (GDA), and artificial neural networks (ANN), demonstrated their power for authentication purposes.