Classification of Montepulciano d'Abruzzo Wines by Linear Discriminant Analysis and Artificial Neural Networks

¹ Dipartimento di Scienze, Università degli Studi, Viale Pindaro, 42, 65127 Pescara, Italy
² Institute of Food Chemistry, Università degli Studi, Via S.Costanzo, 06100, Perugia, Italy
³ A.R.S.S.A. Agenzia Regionale Servizi Sviluppo Agricolo, Piazza Torlonia, Avezzano, Italy.

dapi{at}unipg.it

The data relative to the chemical composition of Montepulciano d'Abruzzo wines were elaborated by an Artificial Neural Networks -ANN- analysis procedure. The same data were also elaborated by a multivariate analysis procedure (Linear Discriminant Analysis -LDA-). These procedures were used in an attempt to classify/characterize the wines. The data came from the chemical analysis of 116 wine samples produced during two years in three zones of Montepulciano d'Abruzzo hypothesized to be different pedoclimatically. Classification of the samples according to the year of production only, as well as according to the pedoclimatic zone only, was attempted. The results show that the analysis allows information to be obtained which is particularly useful for characterizing the wines according to the year of production.

Key words: wine, chemical composition, linear discriminant analysis, artificial neural networks, classification/characterization

Submitted on December 10, 1999
Revised on March 17, 2000