Exploratory analysis of South America wines using artificial intelligence

Abstract

Abstract The microwave-induced plasma optical emission spectrometer has been used for multielement determination in South American wine samples. The analytes were determined after acid digestion of 47 samples of Brazilian and Argentine wine and the application of logistic regression, support vector machine, and decision tree for exploratory analysis and the comparison of these algorithms in differentiating red wine samples by region of origin. The Brazilian wines samples were clustering in a group, while Argentines wines samples formed another cluster. The quantification limits (mg L-1) were P (0.06); B (0.08); K (0.17); Mn (0.002); Cr (0.002) and Al (0.02). The accuracy of the method was evaluated analyzing the wine samples by ICP OES for results comparison. The concentration in mg L-1 found for each element in wine samples, were Al (< 0.02 - 1.82), Cr (0.15 - 0.50), Mn (< 0.002 - 0.8), P (97 - 277), B (1.7 - 11.6), Pb (< 0.06 - 0.3), Na (8.84 - 41.57) and K (604 - 1701), in mg L-1.