The Most Important Parameters to Differentiate Tempranillo and Tempranillo Blanco Grapes and Wines through Machine Learning

A study in order to differentiate Tempranillo and Tempranillo blanco grapes and wines from A.O.C. Rioja (Spain) has been carried out. The three most important groups of chemical compounds in grapes and wines were determined: nitrogen and phenolic compounds by HPLC-DAD, and volatile compounds by GC-MS. A machine learning approach was carried out to achieve the most important chemical parameters to differentiate Tempranillo and Tempranillo blanco grapes and wines based on the Kruskal-Wallis test, F-score feature selector, and support vector machine classification. Four importance levels were established for both grape and wine discrimination. The first level is composed of variables that can differentiate Tempranillo and Tempranillo blanco grapes and wines by analyzing the chemical concentrations. The second, third, and fourth level of importance is composed of a variable set that can correctly classify the samples, each one determined according to the F-score importance ranking and the SVM discriminations. Some chemical compounds demonstrated to be good predictors to wine and grape discriminations, whereas some chemicals are only a good predictor to one discrimination approach.

Automated summary

A study was conducted to differentiate Tempranillo and Tempranillo blanco grapes and wines from A.O.C. Rioja, Spain. Chemical compounds in the grapes and wines were analyzed using HPLC-DAD and GC-MS, with a machine learning approach employed to determine the most important parameters for discrimination. Four importance levels were established, with some chemical compounds showing good predictive capabilities for discrimination.