Study of preferences for white and red wine using binary classification methods

Abstract

The application of data mining methods allows us to detect a series of patterns that may exist in the data we analyze but are not easy to detect a simple view. In this case, we apply some techniques to predict the frequencies of the taste of wine from a series of physical - chemical characteristics of its composition, both wine and white wine, drinks that have been liked by many people internationally for a long time The data set that was used in this work was taken from Green Wine from the North of Portugal. These data had a group of variables that allowed applying classification methods to predict the flavor specifications of the wine based on the criteria given by the customers. The methods were used to achieve this objective: discriminant analysis, logistic regression and neural networks. The results showed that for the two data sets the results are very similar when the three specific methods are applied. The discriminant capacity of the models makes it possible to clearly distinguish the separation of the two groups for classification.