Compute precision–recall curve on imbalanced data

Part 1 — Metric Priority Report for Imbalanced Binary Classification You are given the true binary labels of a dataset and the positive-class probabilities produced by a classifier, together with a decision threshold. Build a deterministic reporting function that measures how lopsided the class mix is, compares the accuracy obtained at the given threshold with the accuracy of the majority-class baseline (that is, max(n_pos, n_neg) / n), and states which families of…