Statistical Analysis of the Performance of Rank Fusion Methods Applied to a Homogeneous Ensemble Feature Ranking

Table 2

The performance average of the rank fusion algorithms based on AUFC criterion for all datasets.

Dataset	Fusion
	Epileptic	USPS	UJI	HAR	Isolet	COIL20
B-Geom	0.820 ± 0.019^(a)	0.847 ± 0.039^(d)	0.909 ± 0.035^(c)	0.906 ± 0.067^(d)	0.759 ± 0.114^(c)	0.814 ± 0.176^(def)
B-L2	0.821 ± 0.018^(a)	0.833 ± 0.039^(a)	0.854 ± 0.053^(a)	0.867 ± 0.076^(a)	0.747 ± 0.109^(a)	0.810 ± 0.181^(abcdef)
B-Mean	0.821 ± 0.019^(a)	0.840 ± 0.037^(c)	0.888 ± 0.040^(f)	0.886 ± 0.067^(c)	0.752 ± 0.111^(b)	0.809 ± 0.180^(cd)
B-Median	0.830 ± 0.021^(b)	0.843 ± 0.040^(abc)	0.913 ± 0.031^(c)	0.905 ± 0.068^(d)	0.756 ± 0.114^(c)	0.802 ± 0.178^(abc)
B-Min	0.849 ± 0.015^(c)	0.866 ± 0.040^(e)	0.914 ± 0.025^(c)	0.941 ± 0.036^(f)	0.777 ± 0.117^(d)	0.830 ± 0.187^(abcdef)
Kwik	0.829 ± 0.018^(b)	0.844 ± 0.040^(abc)	0.915 ± 0.027^(c)	0.904 ± 0.067^(d)	0.754 ± 0.115^(b)	0.800 ± 0.181^(abc)
MC	0.820 ± 0.018^(a)	0.846 ± 0.037^(bd)	0.893 ± 0.038^(e)	0.890 ± 0.064^(b)	0.750 ± 0.111^(ab)	0.802 ± 0.184^(af)
RRA	0.827 ± 0.022^(ab)	0.840 ± 0.035^(bcd)	0.873 ± 0.056^(b)	0.893 ± 0.066^(bc)	0.751 ± 0.096^(bc)	0.818 ± 0.171^(be)
Stuart	0.822 ± 0.018^(a)	0.842 ± 0.036^(bc)	0.901 ± 0.039^(d)	0.900 ± 0.067^(e)	0.755 ± 0.105^(bc)	0.816 ± 0.171^(de)
STDEV	0.011	0.014	0.026	0.024	0.015	0.026