Formula used
Accuracy = correct predictions ÷ all predictions.
Precision = TP ÷ (TP + FP).
Recall = TP ÷ (TP + FN).
Specificity = TN ÷ (TN + FP).
F1 score = 2 × precision × recall ÷ (precision + recall).
How to use
- Paste label pairs or upload a delimited file.
- Identify actual and predicted label columns.
- Choose cleaning, ordering, and normalization settings.
- Select a positive class for binary details.
- Calculate, inspect errors, and export the report.
Example data table
| Actual | Predicted | Predicted2 | Text ID |
|---|---|---|---|
| positive | positive | positive | doc-001 |
| negative | positive | negative | doc-002 |
| neutral | neutral | neutral | doc-003 |
| question | complaint | question | doc-006 |
Frequently asked questions
What does a diagonal cell mean?
It counts records predicted as their correct class.
Why use row normalization?
It compares prediction outcomes within each actual class.
Why use column normalization?
It shows each predicted class composition by actual labels.
What is macro F1?
It averages class F1 scores without support weighting.
What is weighted F1?
It weights each class F1 using its support.
When is balanced accuracy helpful?
It helps when class sizes differ substantially.
Can this compare two models?
Yes. Provide a second predicted-label column.
Does it support multilabel text?
Yes. Enable multilabel mode and set a separator.
How is PDF export handled?
Use Print or Save PDF in your browser.