How to use
- Enter TP, FP, TN, and FN from one binary classification result.
- Optionally rename the positive and negative labels to match your dataset or workflow.
- Read precision beside recall so you can discuss false alarms and missed positives at the same time.
Wave 6 classification metrics
Positive-class trade-off at one threshold
This page stays focused on one operating point. It is narrower than a confusion matrix page and lighter than a threshold-sweep page, so you can explain the positive-class trade-off quickly.
Inputs
Run a calculation to review precision, recall, and F1 from one binary result set.
Precision and recall answer different questions
Precision asks whether predicted positives are trustworthy. Recall asks whether actual positives are being found. If you lower a threshold, recall usually rises first, but precision often falls because more borderline negatives get classified as positive.
When to stay on this page
Stay here when your team already picked one operating threshold and now needs a plain-language readout of the resulting trade-off. This is often the right page for model review slides, screening policy discussions, or threshold sign-off notes.
When to move to ROC AUC
ROC AUC is the next step when the threshold itself is still negotiable. Use that page first if your main question is how sensitivity and specificity move across the full score sweep rather than at one chosen cutoff.
Frequently asked questions
When should I use this page instead of ROC AUC?
Use this page when one threshold is already chosen and you want to inspect precision, recall, and F1 at that operating point. ROC AUC is for score ranking across many thresholds before you commit to one cutoff.
Why can recall improve while precision gets worse?
A lower threshold usually labels more cases as positive. That catches more actual positives, which raises recall, but it also tends to admit more false positives, which lowers precision.
What does F1 add beyond precision and recall?
F1 is the harmonic mean of precision and recall. It gives one summary number that falls when either precision or recall is weak, so it is useful when you need a compact positive-class metric.
Does the share URL include my counts or labels?
No. The share URL stores only lightweight settings such as decimal places. Counts and custom labels stay in your browser.
Related
Comments (optional)
To reduce load, comments are fetched only when needed.