A Learning Framework of Nonparallel Hyperplanes Classifier

Table 5

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

Dataset	1-v-1	1-v-r	MBSVM	MHNPSVM
	Accuracy (%)	Accuracy (%)	Accuracy (%)	Accuracy (%)
	Time (s)	Time (s)	Time (s)	Time (s)
Iris	96.83 ± 1.75	95.73 ± 3.78	95.00 ± 4.95	96.96 ± 1.12
	0.025	0.014	0.009	0.010
Wine	96.59 ± 1.48	97.72 ± 0.74	94.77 ± 4.07	95.88 ± 2.21
	0.058	0.021	0.028	0.023
Ecoli	87.63 ± 0.81	86.77 ± 0.87	85.72 ± 1.02	86.78 ± 0.75
	0.863	0.522	0.097	0.089
Vowel	54.21 ± 2.24	57.44 ± 3.26	59.42 ± 4.96^*	64.60 ± 3.06^*
	1.459	0.580	0.160	0.172
Glass	94.16 ± 1.84	94.42 ± 4.06	92.80 ± 9.80^*	95.83 ± 1.04^*
	1.037	0.405	0.183	0.105
Vehicle	77.79 ± 2.21	78.22 ± 2.10	77.59 ± 2.16	77.13 ± 1.87
	28.11	10.05	2.96	2.58
Car	86.78 ± 0.50	86.72 ± 0.31	84.09 ± 0.33^*	87.79 ± 0.91^*
	16.042	13.79	5.92	6.05
Segment	91.60 ± 2.428	92.54 ± 2.03	92.68 ± 1.87	93.04 ± 2.01
	28.078	15.26	17.04	17.55
Satimage	91.80 ± 0.81	90.20 ± 1.13	92.40 ± 2.08	91.40 ± 1.49
	60.50	32.29	47.45	45.27
Mean accuracy	86.38	86.64	86.05	87.71
Mean time	15.13	8.10	8.21	7.98

A greater difference between MHNPSVM and MBSVM.