Table 4. Comparison of the quality of risk estimation on the test data via the slope and bias of the calibration curve for all ICS setups (lpICS and enICS, each with preselection), Decision Tree (DT), Naive Bayes, linear and nonlinear (RBF) SVM.

Datasets include Acute inflammation (‘Inflammation’ and ‘Nephritis’ labels), Breast Cancer Diagnosis, Cardiotocography (‘Cardio’), Chronic kidney disease (‘Kidney’) and Indian Liver Patient data (‘Liver’).

	Inflammation		Nephritis		Breast Cancer		Cardio		Kidney		Liver
	Slope	Bias	Slope	Bias	Slope	Bias	Slope	Bias	Slope	Bias	Slope	Bias
lpICS	1	0	1	0	0.89	0.03	0.95	−0.01	0.95	0.05	0.44	0.40
lpICS-pre	1	0	1.02	−0.02	0.89	0.03	0.92	0.01	0.98	0.02	0.55	0.32
enICS	1.01	−0.01	1.02	−0.02	0.96	−0.01	1.04	−0.01	0.95	0.05	0.72	0.21
enICS-pre	1	0.01	1.02	−0.02	0.96	−0.01	1.04	0	0.92	0.05	0.72	0.21
DT	1	0	1	0	0.90	0.01	0.94	0.03	0.91	0.09	0.63	0.27
Naive Bayes	1.15	−0.05	1.03	0.05	0.84	0.05	0.74	0.06	0.84	0.18	0.34	0.61
SVM-lin	1	0	1	0	0.96	0.01	1	0	0.99	0.03	0.72	0.22
SVM-rbf	1	0	1	0	0.94	0	0.88	0.03	0.97	0.03	0.68	0.24