ROC and Precision-Recall Curve Overlay

Two-panel ROC + PR curves overlaying multiple models on one chart for comparison.

Prompt

A two-panel figure showing ROC and Precision-Recall curves for 4 binary classifiers.

Left panel — ROC Curve:
- X-axis: False Positive Rate (0-1)
- Y-axis: True Positive Rate (0-1)
- Diagonal "random" reference line in dashed gray.
- Four curves (one per model): Logistic Regression (AUC=0.78), Random Forest (AUC=0.85), XGBoost (AUC=0.89), Neural Net (AUC=0.91).
- Legend with each model and its AUC value.

Right panel — Precision-Recall Curve:
- X-axis: Recall (0-1)
- Y-axis: Precision (0-1)
- Horizontal "baseline" reference at the positive class prevalence (e.g., 0.20).
- Same four models, now reporting AUPRC: 0.62 / 0.74 / 0.81 / 0.85.

Both panels share consistent line colors per model (navy / teal / amber / coral).

Style: clean academic chart, white background, gridlines every 0.1, sans-serif, optimised for medical / binary-classification publications.

Use in Generator

When to use

For binary classification papers where AUC and AUPRC are reported.

Variations

With operating-point markers

Add a star marker on each curve showing the chosen operating point (threshold tuned on validation), with a callout listing precision / recall / FPR at that point.

Tips

Always include the random / baseline reference line. Without it AUC values lack context.
Use consistent colors across both panels. Switching colors between ROC and PR confuses comparisons.
Report both AUC and AUPRC. Imbalanced datasets are misread by AUC alone.

FAQ

How do I show confidence bands?

Add a translucent band around each curve representing the bootstrapped 95% CI. Reduce line opacity slightly so bands stay readable.