Evaluation metric · BEGINNER LESSON
Precision–recall curve
A precision–recall curve is a plot of precision against recall at different decision thresholds. It shows how a binary classifier’s exactness and completeness change as the cutoff changes.
What it is
A precision–recall curve is a graph used for binary classification, where the task is to separate two classes such as positive and negative. Precision measures the fraction of predicted positives that are correct. Recall measures the fraction of actual positives that are found. The curve plots precision against recall while the decision threshold changes. A threshold is the cutoff used to decide whether a score counts as positive.
[1] A Framework for Benchmarking Pathway Reconstruction ... [3] PrecisionRecallCurve — PyTorch-Ignite v0.5.4 Documentation
How it is built
Models often output scores or confidence values, not just yes/no answers. If the threshold is set low, more cases are labeled positive, which can raise recall but may lower precision because more false positives are included. If the threshold is set high, fewer cases are labeled positive, which can raise precision but may lower recall because more false negatives are missed. By computing precision and recall at many thresholds, the curve shows the tradeoff between these two measures.
[1] A Framework for Benchmarking Pathway Reconstruction ... [3] PrecisionRecallCurve — PyTorch-Ignite v0.5.4 Documentation
What the curve helps you judge
The curve is useful when false positives and false negatives matter differently. It can summarize performance across thresholds better than a single yes/no result. The sources also describe Average Precision, which compresses the curve into one number by combining precisions across recall changes. In pathway reconstruction studies, these curves were used to compare predicted nodes or edges against a gold standard, meaning a known reference set.
[1] A Framework for Benchmarking Pathway Reconstruction ... [2] Table 10.
Limits and cautions
A precision–recall curve depends on the quality of the gold standard and on the scoring system used to rank predictions. In pathway reconstruction, the sources note that comprehensive gold standards are often missing, and some evaluations only use node-level references rather than full edge or pathway structure. The sources also note that curves can be hard to compare when each result is shown in a separate plot. Another limitation is that a binary-data analysis can violate assumptions made by methods such as principal component analysis, which the sources mention as a separate issue in their framework.
SEE THE IDEA
How a precision–recall curve changes with threshold
The classifier assigns confidence values to cases.
A cutoff turns scores into predicted positive or negative labels.
Precision tracks correctness among predicted positives; recall tracks how many true positives are recovered.
Each threshold gives one precision-recall pair, and the plotted pairs form the curve.
Scores are filtered by thresholds to produce precision-recall pairs, and those pairs are plotted as the curve.
The curve is a visual record of the tradeoff between finding more positives and keeping predictions accurate.
TRY IT YOURSELF
Test your understanding
Choose an answer, then check the explanation.
PUT IT TOGETHER
Match each concept to its meaning
Drag a label onto its description, or select a label and then select a description. This also works with a keyboard.
4 pairs to match
Sources and further reading
This general lesson was generated from public source material. It explains a concept; it does not interpret a private paper or provide medical advice.