A Novel Ensemble Method for Imbalanced Data Learning: Bagging of Extrapolation-SMOTE SVM

Table 4

The average results on various dataset using algorithms. ROS means random oversampling while RUS is random undersampling. SMEN is SMOTE-ENN. RF is random forest and ADA is AdaBoostM1.

		BEBS	SMOTE	ROS	RUS	SMEN	SVM	RF	ADA
Fertility	G-means	0.613	0.521	0.446	0.532	0.532	0	0.063	0.424
	F-scores	0.285	0.233	0.194	0.23	0.242	0	0.029	0.202
	Precision	0.624	0.159	0.128	0.145	0.171	0	0.05	0.194
Glass 7	G-means	0.923	0.912	0.902	0.906	0.891	0.873	0.914	0.906
	F-scores	0.736	0.882	0.878	0.845	0.877	0.857	0.876	0.835
	Precision	0.902	0.906	0.932	0.842	0.914	0.96	0.914	0.831
Parkinson	G-means	0.792	0.523	0.464	0.392	0.436	0.278	0.791	0.776
	F-scores	0.63	0.442	0.324	0.284	0.336	0.15	0.729	0.681
	Precision	0.671	0.727	0.692	0.278	0.735	0.769	0.809	0.736
Pima	G-means	0.753	0.071	0	0	0	0	0.642	0.685
	F-scores	0.681	0.011	0	0.02	0	0	0.55	0.613
	Precision	0.72	0.654	0	0.07	0	0	0.666	0.632
Segmentation 1	G-means	0.856	0.297	0	0.103	0.262	0	0.873	0.921
	F-scores	0.552	0.143	0	0.024	0.144	0	0.85	0.87
	Precision	0.731	0.871	0	0.143	0.869	0	0.95	0.85
Segmentation 3	G-means	0.884	0.261	0	0.072	0.193	0	0.746	0.833
	F-scores	0.611	0.16	0	0.021	0.094	0	0.665	0.762
	Precision	0.792	0.767	0	0.143	0.667	0	0.825	0.841