Events and Meetings of Italian Statistical Society, Advances in Latent Variables - Methods, Models and Applications

Font Size: 
Evaluating the Italian wine quality with the CRAGGING approach
Marika Vezzoli

Last modified: 2013-06-14

Abstract


Assessing the wine quality is a challenging task due to its multifaceted nature. Indeed, subjective evaluations and objective features are mixed together in order to get fair judgements and effective ranking of wines. To asses wine quality, chemical and sensory tests are commonly used. These tests usually collect a relatively high number of variables among which some certainly play a key role. It is then extremely important to identify these attributes, since they can lead to significant improvements in the understanding process of the wine quality. Using a data mining approach, we inspect the quality of Italian red and white wines trying to identify which of the sensorial and chemical-type variables have a major impact on it. In detail, we analyze the data set used by Altroconsumo, an Italian independent consumer's association, for the Guida Vini 2011, containing 231 wines grouped with respect to the type of grapes used by producers. Since the data set has a hierarchical structure, we use anew algorithm, called CRoss-validation AGGregatING (CRAGGING).Moreover, we extract a synthetic model able to identify the predictors and correspondent thresholds useful for showing the "true path" towards the quality.

Full Text: PDF