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

Table 6

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

PIMA		Baseline	Tml	RUS	ENN	HDUS
KNN	Sensitivity (%)	60.06	64.52	70.77	70.57	83.87
	Specificity (%)	83.85	80	73.85	74.92	62.50
	Precision (%)	63.16	56.61	56.41	56.46	50.35
	F1_m (%)	61.57	60.31	62.78	62.73	62.92
	Bacc (%)	70.96	72.26	72.41	72.75	73.19
SVM	Sensitivity (%)	0	66.13	74.42	74	79.03
	Specificity (%)	100	83.85	73.85	76.92	71.50
	Precision (%)	0	58.13	58.54	58.04	57.00
	F1_m (%)	0	61.87	65.53	65.06	66.23
	Bacc (%)	50	74.99	74.63	75.86	75.27
DT	Sensitivity (%)	61.29	69.35	70.97	70.97	91.90
	Specificity (%)	79.23	66.92	58.46	68.46	66.77
	Precision (%)	58.46	50	44.9	50.76	56.98
	F1_m (%)	59.84	58.11	55	59.19	70.34
	Bacc (%)	70.26	68.14	64.71	69.71	79.34
AVG	Sensitivity (%)	39.78	66.67	72.05	71.85	84.93
	Specificity (%)	87.69	76.92	68.72	73.43	66.92
	Precision (%)	40.54	54.91	53.28	54.09	54.78
	F1_m (%)	40.16	60.22	61.26	61.72	66.50
	Bacc (%)	63.74	71.8	70.58	72.77	75.93