Name: Imbalanced Data Lab
Author: MySimulator

Dataset

Minority class share: 5%

Fix

The minority class (fraud, disease, defect — the rare event you actually care about) is what "imbalanced classification" is about. A model that just predicts the majority class every time can still score very high accuracy.

Held-out test metrics

Accuracy—

Minority recall—

Minority precision—

Minority F1—

Confusion matrix (test set)