Area Under The PR Curve measures the area under a precision-recall curve for a classification model. It summarizes the tradeoff between precision and recall across decision thresholds and is especially useful when positive examples are rare or class imbalance is significant.