精确率、召回率、F1 与 ROC-AUC

When one class dominates, accuracy looks great even for a useless model. You need sharper metrics to judge real performance.

Every prediction is a true positive, true negative, false positive, or false negative. These four counts power every richer metric.

Precision asks: of everything you flagged positive, how many truly were? High precision means few false alarms.

Recall asks: of all the real positives, how many did you catch? High recall means few missed cases.

Push precision up and recall often drops, and the reverse. The right balance depends on the cost of each kind of mistake.

The F1 score is the harmonic mean of precision and recall. It rewards models that keep both reasonably high.

A classifier outputs a probability; the threshold decides positive versus negative. Sliding it reshapes precision and recall.

The ROC curve plots true positive rate against false positive rate across every threshold, showing the full tradeoff at a glance.

ROC-AUC is the area under that curve. It captures ranking quality with one threshold-free number from zero to one.

An AUC of 0.5 is random guessing, while 1.0 is perfect separation. Higher means the model ranks positives above negatives well.

Choose your metric by the task: recall for disease screening, precision for spam filters, AUC for ranking quality.

Think about which metric counts how many real positives you successfully caught.

Go beyond accuracy with precision, recall, F1, and ROC-AUC. Each answers a different question, so match the metric to your goal. 🎯