A Learning Framework of Nonparallel Hyperplanes Classifier

Table 6

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

Dataset	1-v-1	1-v-r	MBSVM	MHNPSVM
	Accuracy (%)	Accuracy (%)	Accuracy (%)	Accuracy (%)
	Time (s)	Time (s)	Time (s)	Time (s)
Iris	98.93 ± 1.11	97.63 ± 5.46	98.12 ± 2.08	98.74 ± 1.92
	0.0054	0.0264	0.037	0.030
Wine	97.08 ± 3.32	97.72 ± 0.86	96.45 ± 1.29	97.28 ± 0.96
	7.294	4.6504	0.592	0.523
Ecoli	92.27 ± 1.03	90.35 ± 0.47	91.06 ± 1.45^*	92.95 ± 0.89^*
	0.382	0.0843	0.154	0.182
Glass	98.09 ± 1.04	99.14 ± 0.97	98.76 ± 1.22	99.24 ± 0.93
	0.692	0.1085	0.089	0.092
Vowel	91.37 ± 0.86	94.32 ± 0.18	80.42 ± 4.37^*	85.86 ± 4.72^*
	1.482	0.3844	0.623	0.593
Vehicle	81.03 ± 5.73	82.49 ± 4.26	82.01 ± 1.33	83.57 ± 1.79
	19.562	11.456	2.81	2.50
Car	88.37 ± 0.55	87.36 ± 0.68	85.74 ± 0.33	86.57 ± 0.46
	3.6571	0.9405	1.832	1.944
Segment	95.15 ± 6.02	94.65 ± 4.38	95.96 ± 4.08	95.90 ± 3.29
	128.42	91.69	53.27	49.58
Satimage	93.80 ± 1.46	93.05 ± 1.46	94.03 ± 1.93	94.47 ± 1.58
	190.27	132.47	89.05	88.36
Mean accuracy	92.90	92.97	91.39	92.73
Mean time	39.08	26.87	16.50	15.98

A greater difference between MHNPSVM and MBSVM.