Characteristics of the optimal predictors for the different subtypes.

In each cell the optimal combination of classifier, and feature selection method, is shown.Legend:classifiers: NB = Naive Bayes, NM = Nearest Mean, LREG: Logistic regression, SVM = Support vector machine, 3NN = 3-Nearest Neighbor; Feature selection methods: CFS = Correlated feature selection, WMW = Wilcoxon-Mann-Whitney, BWR = Ratio between to within class sum of squares, WMW-uncor. = Wilcoxon-Mann-Whitney where correlated features are removed, Inf.gain = information gain.