Precision Recall Curves
Move a score threshold and measure precision, recall and stepwise average precision.
Free theory articleMove a score threshold and measure precision, recall and stepwise average precision.
Concept & interpretation
Precision-Recall (PR) curves are used to evaluate the performance of classification models, particularly when classes are highly imbalanced. They illustrate the trade-off between precision (the fraction of positive predictions that are correct) and recall (the fraction of actual positives that are successfully predicted) across different decision thresholds.
Interactive Experiment
You can interactively explore the precision and recall curve for different test datasets using the Precision Recall Curves module. Follow these steps:
- Set the Dataset control to "Four".
- Adjust the Threshold control to 0.35.
- Observe that at this threshold, the model correctly identifies both positive examples (TP = 2), reaching a Recall of 1.0. However, it also incorrectly classifies one negative example as positive (FP = 1), which lowers the Precision to approximately 0.6667.
- Change the Dataset control to "Ties" and note how ties in predicted scores are handled simultaneously, creating a single joint step in the PR curve.