HDEC: A Heterogeneous Dynamic Ensemble Classifier for Binary Datasets

Table 5

The best geometric mean of the ensemble approaches.

Dataset	Best-bagging		Best-AdaBoost		Best-stacking		CSS [36]	HDEC
	Type	Acc.	Type	Acc.	Type (meta)	Acc.
AUS	DT	86.60 ± 1.13	DS	86.60 ± 1.01	BN	86.78 ± 1.27	—	87.05 ± 1.41
BRC	BN	61.00 ± 1.92	DS	58.16 ± 1.95	BN	64.74 ± 0.20	61.57 ± 2.31	65.77 ± 1.06
COL	RT	82.25 ± 0.59	DS	80.35 ± 1.27	BN	83.06 ± 0.09	81.47 ± 1.93	85.85 ± 0.64
DBT	BN	72.48 ± 1.05	BN	71.78 ± 0.95	NB	72.46 ± 0.44	71.44 ± 0.96	75.11 ± 0.51
CLS	KNN	100	KNN	100	All– (K, DsT)	100	—	100
HRT	NB	83.65 ± 0.29	SVM	83.90 ± 0.22	DS	83.90 ± 0.32	—	84.29 ± 0.46
HPT	NB	78.01 ± 0.84	NB	71.92 ± 1.93	BN	76.81 ± 2.46	72.53 ± 5.93	78.35 ± 2.46
ION	RT	91.34 ± 0.42	NB	92.73 ± 0.19	BN	92.18 ± 0.55	—	92.94 ± 0.46
LVD	DT	67.23 ± 2.43	DT	66.61 ± 1.53	NB	62.83 ± 0.73	68.48 ± 1.82	73.46 ± 1.02
SON	K	84.43 ± 0.42	KNN	86.20 ± 0.24	RT	88.77 ± 0.49	—	90.48 ± 0.25
GDE	BN	66.97 ± 1.48	RT	70.74 ± 1.27	NB	75.44 ± 1.37	—	78.12 ± 1.91
VOT	DT	96.27 ± 0.31	NB	95.88 ± 0.52	DS	96.11 ± 0.56	95.80 ± 1.08	96.21 ± 0.29