Fig. 2From: Development of a predictive machine learning model for pathogen profiles in patients with secondary immunodeficiencyReceiver Operating Characteristic (ROC) Curves for Machine Learning Models. A-D Comparison of model performance through ROC analysis. A K-Nearest Neighbors (KNN) ROC curves for various classes, highlighting the trade-off between true positive rate and false positive rate with an area under the curve (AUC) for the micro-average at 0.87 and macro-average at 0.91. B Boosted Logistic Regression ROC curves, showing improved performance with an AUC for the micro-average at 0.93 and macro-average at 0.92. C Random Forest ROC curves, indicating superior performance with an AUC for the micro-average at 0.98 and macro-average at 0.98. D Gradient Boosting Machine ROC curves, exhibiting exceptional discriminative power with an AUC for the micro-average at 0.98 and macro-average at 0.97Back to article page