A Learning Framework of Nonparallel Hyperplanes Classifier

Table 4

Tenfold testing percentage test set accuracy (%) on binary-class UCI datasets for nonlinear classifiers.

Datasets	TWSVM	SVM	GEPSVM	BHNPSVM
	Accuracy %	Accuracy %	Accuracy %	Accuracy %
	Time (s)	Time (s)	Time (s)	Time (s)
Hepatitis	83.39 ± 7.31	84.13 ± 6.25	80.00 ± 5.2	83.40 ± 3.58
	0.016	0.0142	0.0035	0.0697
BUPA liver	67.83 ± 6.49^*	68.32 ± 7.20	63.01 ± 7.46	74.24 ± 0.64^*
	0.033	0.0129	1.305	0.1522
Heart-statlog	82.96 ± 4.67^*	83.33 ± 9.11	86.52 ± 7.36	84.04 ± 4.56^*
	0.029	0.0250	0.438	0.1120
Votes	94.91 ± 4.37	95.64 ± 7.23	94.5 ± 3.37	95.21 ± 5.18
	0.072	0.0495	0.087	0.0152
WPBC	81.28 ± 5.92	80.18 ± 6.90	80.07 ± 5.97	80.89 ± 1.17
	0.029	0.0148	0.0043	0.0468
Sonar	89.64 ± 6.11	88.93 ± 10.43	81.93 ± 4.41	88.05 ± 1.79
	0.014	0.0781	0.020	0.2896
Australian	75.8 ± 4.91^*	85.51 ± 4.85	69.55 ± 5.37	77.58 ± 2.53^*
	0.420	0.0425	0.334	0.497
Pima-Indian	73.74 ± 5.2^*	76.09 ± 3.58	74.66 ± 5.00	77.70 ± 0.39^*
	0.427	0.442	15.892	0.381
CMC	73.95 ± 3.48^*	68.98 ± 3.44	68.67 ± 3.84	78.43 ± 0.13^*
	1.708	1.755	1.042	1.920
Mean accuracy	80.39	81.23	77.66	82.17
Mean time	0.3053	0.27	2.1251	0.3871

A greater difference between BHNPSVM and TWSVM.