Clustering and classification of red wines by physical-chemical properties using data mining methods

A study was preformed of 178 samples of red wines of Italian manufacturers, taken from the public repository for machine learning UCI. The methods of Data Minig made a computer analysis of the influence of 13 physico-chemical properties of the samples on the distribution of wines in three groups. The next classification models were built: factor, discriminant, canonical, MLP multilayer perceptron, Kohonen neural network SOFM, predictive (support vector method, Bayesian klаssifier, nearest neighbor method) and decision trees. The neural network classifiers SOFM 13-3, MLP 13-5-3 and clusterer SOFM 16-3 were trained. It was shown that proline, flavonoids, color intensity, proteins and alcohol determine the discriminatory power of the models.

data mining, statistica, red wines, clustering, classification, data mining, statistica