Fig. 3

Confusion Matrices for Model Performance Evaluation. This figure presents the confusion matrices of four machine learning models, allowing for a detailed assessment of prediction accuracy for various pathogen classes. Values along the diagonal represent correct classifications, while non-diagonal values indicate misclassifications. A The K-Nearest Neighbors (KNN) model matrix, with counts of true vs. predicted labels, showing a specific number for true positive rates in pathogen detection. B The Boosted Logistic Regression model matrix, detailing the true positives along the diagonal and misclassifications off-diagonal, reflecting the model’s predictive power and misclassification patterns. C The Random Forest model matrix, which illustrates a higher concentration of true positives along the diagonal, indicative of a model with strong predictive accuracy. D The Gradient Boosting Machine model matrix, showing high true positive rates, especially for certain pathogens, suggesting a high degree of model precision