| Popis: |
In connection with the growing consumer’s interest to Russian wines with controlled place of origin PGI and PDO, the most pressing issue is the method of their identification. One of the most effective ways to confirm the wine's place of origin in world practice is a comprehensive research of the elemental profile and isotopic characteristics of “light” elements using the methods of statistical analysis. We have selected 32 samples of fresh grapes from various wine regions of Russia (Krasnodar Territory, Republic of Crimea, Republic of Dagestan). The grape must obtained from them was fermented under laboratory conditions. In the prepared wines, the elemental profile was determined, which included 71 indicators, as well as indicators δ18О, δD of released ethanol and δ18О of the wine water. The resulting data set was analyzed using statistical methods PCA, Permanova, the Mann-Whitney test, and machine learning was also performed. It is shown that the difference between the values of the mass concentration of the elements Al, Fe, Br, Re, U for samples from Krasnodar Territory and the Republic of Crimea are statistically significant. On the matrix of the obtained values, the Random Forest model was trained, which was able to distinguish the regions of wine origin with an accuracy of 90%. When analyzing the nonlinear dependence, the indicators of Si, Li, Co, Cu, Ba, Na, Ni, U, Al, S, Fe, Mn, B and δ18О of the water were determined by the model as important. |
