Journal of the Brazilian Computer Society

Table 5 Quantification error and statistical analysis

From: Combining instance selection and self-training to improve data stream quantification

	SQSI with RF						SQSI with SVM
	CC (NRAE =0.0648)		PCC (NRAE =0.0546)		ACC (NRAE =0.0599)		CC (NRAE =0.0574)		PCC (NRAE =0.0392)		ACC (NRAE =0.0504)
	NRAE	p value	NRAE	p value	NRAE	p value	NRAE	p value	NRAE	p value	NRAE	p value
SQSI-IS +k-means	0.0758	ns	0.0670	ns	0.0846	ns	0.1059	ns	0.0758	ns	0.0853	ns
SQSI-IS + Farthest-first	0.1053	ns	0.0952	ns	0.1102	ns	0.1000	ns	0.0907	ns	0.1165	ns
SQSI-IS + Random	0.1011	ns	0.0867	ns	0.0966	ns	0.1060	ns	0.0953	ns	0.0967	ns

Mean error and results for Friedman test between the SQSI and SQSI-IS using different instance selection techniques for each quantification method
ns not significant

Back to article page