Classification Evaluation Metrics
Accuracy is meaningless under class imbalance; precision and recall trade off false positives against false negatives, F1 combines them, and ROC-AUC and average precision evaluate the ranking across all thresholds. Choose the metric from the cost structure of the application first, then optimize — the threshold is a business decision, not 0.5 by default.
This Concept is waiting for its first lesson!
Accuracy is meaningless under class imbalance; precision and recall trade off false positives against false negatives, F1 combines them, and ROC-AUC and average precision evaluate the ranking across all thresholds. Choose the metric from the cost structure of the application first, then optimize — the threshold is a business decision, not 0.5 by default.
Are you a teacher? Sign in to start contributing.
Sign In