BMC Medical Informatics and Decision Making

Table 4 Evaluation the effect of moving threshold and weighing in performance of the algorithms

From: Diabetes mellitus risk prediction in the presence of class imbalance using flexible machine learning methods

	Accuracy	F1-measure	G-mean	MCC	AUROCC	AUPRC
DNN
g-t	0.784	0.554	0.786	0.732	0.857	0.603
f1-t	0.848	0.591	0.757	0.750	0.857	0.603
weighted	0.822	0.581	0.780	0.744	0.858	0.606
XGBoost
g-t	0.774	0.538	0.774	0.721	0.854	0.622
f1-t	0.855	0.586	0.738	0.749	0.854	0.622
weighted	0.832	0.588	0.776	0.748	0.853	0.620
Random forest
g-t	0.777	0.534	0.767	0.717	0.840	0.578
f1-t	0.841	0.564	0.733	0.734	0.840	0.578
weighted	0.810	0.566	0.775	0.735	0.846	0.591

g-t maximum g-mean based moved threshold, f1-t maximum f1-measure based moved threshold

Back to article page

ISSN: 1472-6947

Contact us

General enquiries: journalsubmissions@springernature.com