Development of a machine learning prediction model for loss to follow-up in HIV care using routine electronic medical records in a low-resource setting

Table 2 Performance of ML algorithms on original imbalanced data vs. data balanced with class balancing and SMOTE

ML algorithm	Comparison metrics	Original imbalanced classes (%)	Class balancer (%)	SMOTE (%)
RF	Accuracy	84.8	84.1	84.2
	Sensitivity	68.3	82.3	82.4
	AUC	89.1	89.0	89.5
J48	Accuracy	85.2	83.7	83.9
	Sensitivity	66.8	82.2	82.5
	AUC	87.2	87.8	88.0
K-NN	Accuracy	84.9	84.0	84.2
	Sensitivity	66.8	82.3	82.4
	AUC	89.1	89.1	89.5
SVM	Accuracy	85.0	81.8	81.9
	Sensitivity	66.3	80.1	80.0
	AUC	79.8	81.8	81.8
LR	Accuracy	84.6	81.1	81.7
	Sensitivity	68.2	80.3	78.1
	AUC	88.6	88.5	88.5
Naïve Bayes	Accuracy	83.9	81.2	81.8
	Sensitivity	70.9	79.3	75.9
	AUC	88.3	88.3	88.3

Abbreviations: AUC - area under the curve, J48 is a decision tree algorithm based on the C4.5 algorithm (J48 is its implementation in Weka). K-NN = k-nearest neighbors, LR = logistic regression, ML = machine learning, RF = random forest, SMOTE = synthetic minority oversampling technique, SVM = support vector machine

ISSN: 1472-6947