[Retracted] The Use of Hellinger Distance Undersampling Model to Improve the Classification of Disease Class in Imbalanced Medical Datasets

Table 7

Evaluation results for THS dataset using classifiers (KNN, SVM, and DT) and models (baseline, Tml, RUS, ENN, and HDUS).

THS		Baseline	Tml	RUS	ENN	HDUS
KNN	Sensitivity (%)	0.00	0.00	42.86	4.76	23.81
	Specificity (%)	100.00	98.97	60.82	91.75	75.26
	Precision (%)	0.00	0.00	19.15	11.11	19.24
	F1_m (%)	0.00	0.00	26.47	6.67	21.28
	Bacc (%)	50.00	49.48	51.84	48.26	49.53
SVM	Sensitivity (%)	0.00	0.00	66.67	4.76	71.43
	Specificity (%)	100.00	100.00	47.42	91.75	44.27
	Precision (%)	0.00	0.00	21.54	11.11	21.13
	F1_m (%)	0.00	0.00	32.56	6.67	32.61
	Bacc (%)	50.00	50.00	57.04	48.26	57.84
DT	Sensitivity (%)	23.81	14.29	42.86	38.10	80.95
	Specificity (%)	87.63	91.75	48.45	81.44	40.02
	Precision (%)	29.41	27.27	15.25	30.77	25.99
	F1_m (%)	26.32	18.75	22.50	34.04	39.33
	Bacc (%)	55.72	53.02	45.66	59.77	60.48
AVG	Sensitivity (%)	7.94	4.76	50.79	15.87	58.73
	Specificity (%)	95.88	96.91	52.23	88.32	53.18
	Precision (%)	9.80	9.09	18.65	17.66	22.12
	F1_m (%)	8.77	6.25	27.18	15.79	31.07
	Bacc (%)	51.91	50.83	51.51	52.09	55.95