Table 13.

Numerical results of the shallow models of Table 12 when they are used to classify the Haralick and Haralick-plus-Zernicke input features extracted from the CP and NCP scans of the test dataset of Table 3.

Model	Accuracy	Precision	Recall	F-measure	AUC	Training time
Haralick input features
SVM-2D	93.00	0.9386	0.9300	0.9343	0.9940	1.6
SVM-3D	90.20	0.9181	0.9200	0.9190	0.9910	1.8
SVM-RBF	91.20	0.9252	0.9120	0.9186	0.9700	1.5
MLP-50	61.60	0.6160	0.6160	0.6159	0.6600	6.4
MLP-100	70.10	0.8083	0.7010	0.7508	0.8000	9.3
MLP-200	76.10	0.7947	0.7610	0.7775	0.7800	12.2
RF-100	68.50	0.7652	0.6850	0.7229	0.8500	6.4
RF-500	68.70	0.7665	0.7210	0.7431	0.9600	31.8
RF-1000	68.90	0.7676	0.6890	0.7262	0.9620	63.5
Haralick-plus-Zernicke input features
SVM-2D	93.40	0.9417	0.9340	0.9378	0.9950	1.7
SVM-3D	92.70	0.9363	0.9270	0.9316	0.9930	2.4
SVM-RBF	91.70	0.9788	0.9170	0.9469	0.9600	1.6
MLP-50	76.00	0.8132	0.7600	0.7857	0.7900	7.5
MLP-100	77.40	0.7874	0.7740	0.7806	0.7200	10.8
MLP-200	78.30	0.8123	0.7830	0.7974	0.7800	13.2
RF-100	71.10	0.7852	0.7110	0.7463	0.9600	7.3
RF-500	72.00	0.7903	0.7200	0.7535	0.9650	36.5
RF-1000	72.10	0.7893	0.7210	0.7536	0.9680	74.0

Accuracy is in percentage while the training time is measured in seconds.