ROC Curve Tool

Explore ROC points, AUC, and threshold diagnostics. Review classifier performance using sensitivity and specificity trends. Find balanced cutoffs faster with intuitive visuals and exports.

Enter classifier results

Paste two columns in the dataset box: actual label and predicted score. Use the delimiter you select below.

Example format: 1,0.92 on each line. A header row is allowed when the second field is non-numeric.

Example data table

This sample illustrates how labels and prediction scores should be organized before analysis.

# Actual label Predicted score Interpretation
110.98Very strong positive signal
210.94Strong positive signal
310.81Likely positive case
400.61Risk of false positive
510.54Borderline positive score
600.46Borderline negative score
700.29Likely negative case
810.12Missed positive candidate

Formula used

Sensitivity or TPR = TP / (TP + FN)

Specificity = TN / (TN + FP)

False positive rate = FP / (FP + TN) = 1 - Specificity

Precision = TP / (TP + FP)

F1 score = 2 × Precision × Recall / (Precision + Recall)

Youden index = Sensitivity + Specificity - 1

Balanced accuracy = (Sensitivity + Specificity) / 2

AUC is the trapezoidal area under the ROC curve.

The ROC curve is created by recalculating confusion-matrix metrics across every unique threshold in your score column.

How to use this calculator

  1. Paste your actual labels and predicted scores into the dataset box.
  2. Choose the delimiter, positive class label, and score direction.
  3. Leave threshold blank to evaluate the best Youden cutoff automatically.
  4. Submit the form to generate AUC, ROC points, and confusion details.
  5. Inspect the graph, best thresholds table, and selected threshold metrics.
  6. Use the CSV or PDF buttons to export your analysis report.

FAQs

1. What does the ROC curve show?

It shows how true positive rate changes against false positive rate across many decision thresholds. This reveals how well a scoring model separates positive and negative classes.

2. What is a good AUC value?

Higher AUC values indicate stronger ranking performance. Around 0.5 suggests random separation, while values closer to 1.0 indicate excellent discrimination between classes.

3. Why does the tool calculate Youden index?

Youden index helps identify a balanced threshold by maximizing sensitivity plus specificity. It is useful when false positives and false negatives carry similar importance.

4. Can I use labels other than 0 and 1?

Yes. Enter any positive class label you want, such as yes, disease, fraud, or pass. The tool matches that value exactly during analysis.

5. What if lower scores mean positive cases?

Choose the lower-score option in the form. The calculator will reverse the threshold rule and build the ROC curve correctly for that scoring direction.

6. Does this tool need probabilities?

No. It works with any continuous or ordinal score, as long as larger or smaller values consistently indicate stronger evidence for one class.

7. Why are there threshold values like INF or -INF?

Those boundary thresholds create the ROC endpoints. They represent the cases where no observations are predicted positive or where all observations are predicted positive.

8. What should I export, CSV or PDF?

Use CSV when you want rows for spreadsheet analysis. Use PDF when you need a printable summary with key metrics, confusion counts, and the ROC chart.

Related Calculators

prevalence calculatorhistogram generatorhedges g calculatorglass deltalinear regression tooldiagnostic accuracy calculatorpublication bias testsample mean estimator

Important Note: All the Calculators listed in this site are for educational purpose only and we do not guarentee the accuracy of results. Please do consult with other sources as well.