ROC Curves in Python: How to Plot AUC, Thresholds, and Confidence Intervals
ROC curves are useful when you need to show how well a classifier separates two classes across all thresholds. In clinical research, they are often paired with AUC, sensitivity, specificity, and an operating threshold chosen for a practical cutoff rather than a mathematically perfect one.
AUC
Summarizes ranking performance across thresholds.
Threshold
Shows where sensitivity and specificity balance for your use case.
Confidence
Useful when reviewers expect uncertainty around model performance.
Compact ROC code
For most publication figures, start with `roc_curve` and `roc_auc_score`, then annotate the threshold you plan to use in the real workflow.
Try it
Try it now: generate your ROC curve
Build an ROC figure with AUC annotation and threshold-aware styling from your own classification results.
Generate my ROC curve →Newsletter
Get a weekly Python plotting tip
One concise tip each week for cleaner, faster scientific figures. Built for researchers who publish.
from sklearn.metrics import roc_curve, roc_auc_score
import matplotlib.pyplot as plt
import numpy as np
y_true = np.array([0, 0, 1, 1, 0, 1, 0, 1])
y_score = np.array([0.10, 0.35, 0.72, 0.81, 0.22, 0.90, 0.40, 0.76])
fpr, tpr, thresholds = roc_curve(y_true, y_score)
auc = roc_auc_score(y_true, y_score)
fig, ax = plt.subplots(figsize=(6, 6))
ax.plot(fpr, tpr, lw=2, label=f"Model (AUC = {auc:.2f})")
ax.plot([0, 1], [0, 1], "--", color="gray", label="Chance")
ax.set_xlabel("False Positive Rate")
ax.set_ylabel("True Positive Rate")
ax.set_title("ROC Curve")
ax.legend(frameon=False)
plt.tight_layout()What to label
- State the AUC in the legend or caption.
- Mark the chosen operating point if the figure supports a clinical decision.
- Explain whether the curve is from cross-validation, holdout data, or a test set.
- If you show a confidence band, say how it was computed.
Good fit
Diagnostic tests, binary classifiers, biomarker studies, and any scenario where a threshold decision matters.
Poor fit
Problems with heavily imbalanced classes, multi-class workflows, or cases where precision-recall curves communicate the story more honestly.
If you need a broader clinical workflow, read the clinical research figures guide, then compare it with the ROC technique page.
Try Plotivy with classifier resultsRelated chart guides
Apply this tutorial directly in the chart gallery with ready-to-run prompts and examples.
Technique guides scientists read next
scipy.signal.find_peaks guide
Tune prominence and width parameters for robust peak extraction.
Savitzky-Golay smoothing
Reduce noise while preserving peak shape and position.
PCA visualization workflow
Move from high-dimensional measurements to interpretable components.
ANOVA with post-hoc brackets
Add statistically correct pairwise significance annotations.
Found this helpful? Share it with your network.
Experimental Physicist & Photonics Researcher
Hands-on experience in silicon photonics, semiconductor fabrication (DRIE/ICP-RIE), optical simulation, and data-driven analysis. Built Plotivy to help researchers focus on discoveries instead of data struggles.
More about the authorVisualize your own data
Apply the techniques from this article to your own datasets. Upload CSV, Excel, or paste data directly.