Using data mining for wine quality assessment

Abstract

Certification and quality assessment are crucial issues within the wine industry. Currently, wine quality is mostly assessed by physico- chemical (e.g alcohol levels) and sensory (e.g. human expert evaluation) tests. In this paper, we propose a data mining approach to predict wine preferences that is based on easily available analytical tests at the certifi- cation step. A large dataset is considered with white vinho verde samples from the Minho region of Portugal. Wine quality is modeled under a re- gression approach, which preserves the order of the grades. Explanatory knowledge is given in terms of a sensitivity analysis, which measures the response changes when a given input variable is varied through its do- main. Three regression techniques were applied, under a computationally efficient procedure that performs simultaneous variable and model selec- tion and that is guided by the sensitivity analysis. The support vector machine achieved promising results, outperforming the multiple regres- sion and neural network methods. Such model is useful for understand- ing how physicochemical tests affect the sensory preferences. Moreover, it can support the wine expert evaluations and ultimately improve the production.