Table 7.

A comparison between FRaC with and without discretizing numeric features

Feature predictor	Discrete FRaC	FRaC	p-value
Decision/regression tree	9	25	0.000528
Linear Kernel SVM	10	27	0.000244
RBF Kernel SVM	12	26	0.00123
Tree, linear and RBF SVM combined	9	26	0.00146

Columns have the same meaning as in Table 4 (Rows for naïve Bayes and RIPPER are not shown because these algorithms are not compatible with numeric targets and therefore the ability of FRaC to handle numeric features is not an advantage when using these models)