kuva pr

Precision-recall curve — evaluates binary classifiers when class imbalance makes ROC curves optimistic.

Input: one row per sample with a numeric score column and a binary label column (1 = positive, 0 = negative).

FlagDefaultDescription
--score-col <COL>0Classifier score column (higher = more positive)
--label-col <COL>1True label column (1/0 or true/false)
--color-by <COL>Group column; one curve per unique value
--no-baselineoffHide the no-skill (prevalence) baseline
--auc-labeloffAppend AUC-PR value to each curve's legend entry
--legend <LABEL>Add a legend
kuva pr data.tsv --score-col score --label-col label --auc-label

kuva pr data.tsv --score-col score --label-col label \
    --color-by model --auc-label --legend "Model" \
    --title "Precision-Recall Curves"

See also: Shared flags — output, appearance, axes.