This data set consists of the results of a chemical
analysis of wines derived from three different cultivars [37,
38]. There are 13 continuous features, with no missing values.
There are 59, 71, and 48 members of each of the three classes,
respectively. The three classes are nearly linearly separable,
and linear discriminant analysis can obtain 98.9% accuracy
over all three classes. This data set is thus better for evaluating
feature selection capability than classifier accuracy.