An adaptive optimal ensemble classifier via bagging and rank aggregation with applications to high dimensional data

Table 1 Threenorm simulation data

	Accuracy	Sensitivity	Specificity	AUC
SVM	0.451900	0.468200	0.435600	0.429016
	(0.00988)	(0.02144)	(0.02314)	(0.01318)
RF	0.562200	0.557600	0.566800	0.591170
	(0.00540)	(0.00853)	(0.00806)	(0.00635)
PLS + LDA	0.610000	0.608000	0.612000	0.610032
	(0.00561)	(0.00860)	(0.00797)	(0.00561)
PCA + LDA	0.503600	0.501800	0.505400	0.505236
	(0.00617)	(0.00674)	(0.00680)	(0.00753)
PLS + RF	0.612200	0.586400	0.638000	0.648102
	(0.00506)	(0.01250)	(0.01198)	(0.00595)
PLS + QDA	0.607500	0.617200	0.597800	0.607500
	(0.00577)	(0.01142)	(0.01218)	(0.00577)
PLR	0.540800	0.538000	0.543600	0.557342
	(0.00459)	(0.00819)	(0.00804)	(0.00553)
PLS	0.600300	0.600400	0.600200	0.647896
	(0.00542)	(0.01319)	(0.01361)	(0.00609)
Greedy	0.596600	0.581800	0.611400	0.621590
	(0.00559)	(0.01117)	(0.01045)	(0.00657)
Ensemble	0.613000	0.606200	0.619800	0.653700
	(0.00563)	(0.00823)	(0.00729)	(0.00587)

Average accuracy, sensitivity, specificity and AUC for 100 datasets from the threenorm data with N = 100 and d = 1000. Standard errors are reported in parentheses.

ISSN: 1471-2105