Machine learning to predict virological failure among HIV patients on antiretroviral therapy in the University of Gondar Comprehensive and Specialized Hospital, in Amhara Region, Ethiopia, 2022

Table 3 compares imbalanced data handling techniques using accuracy and Area under the curve (AUC)

Algorithms	Comparison method	Unbalanced	SMOTE	Under-Sampling	ADASYN
Logistic Regression	Accuracy (%)	97.23	95.11	82.94	94.40
Logistic Regression	AUC	0.915	0.987	0.913	0.986
K Nearest Neighbours	Accuracy (%)	97.34	95.54	81.38	93.20
K Nearest Neighbours	AUC	0.705	0.986	0.881	0.973
Decision Tree	Accuracy (%)	95.82	98.08	81.83	93.34
Decision Tree	AUC	0.619	0.982	0.817	0.935
Random Forest	Accuracy (%)	97.42	98.80	83.34	95.10
Random Forest	AUC	0.892	0.999	0.917	0.991
Gradient Boosting	Accuracy (%)	97.17	95.22	84.52	93.06
Gradient Boosting	AUC	0.903	0.988	0.912	0.982
XGBoost	Accuracy (%)	96.98	95.25	82.58	96.24
XGBoost	AUC	0.870	0.997	0.901	0.994
Support Vector Machine	Accuracy (%)	97.32	96.99	84.60	95.19
Support Vector Machine	AUC	0.867	0.995	0.940	0.991

AUC Area Under Curve, SMOTE Synthetic Minority Over-sampling Technique, ADASYN Adaptive Synthetic. Underline and bold numbers were the highest score of the classifier

ISSN: 1472-6947