Overview
While the ROC curve is popular, the Precision-Recall (PR) curve is often more informative when the 'positive' class is very rare (e.g., detecting a rare disease or credit card fraud).
Interpretation
A high area under the PR curve represents both high recall and high precision. A perfect model would hug the top-right corner.
Comparison with ROC
ROC curves can be misleadingly optimistic on imbalanced datasets because they include True Negatives in the calculation. PR curves focus only on the positive class, providing a tougher and more realistic evaluation.