Skip to content
Sign in

What is ROC Curve?

A ROC Curve is a graph that illustrates the performance of a binary classification model by plotting the true positive rate against the false positive rate at various decision thresholds.

It is created by varying the classification threshold from 0 to 1 and calculating the rates of correctly and incorrectly identified positive cases for each value. The resulting curve shows the trade-off between sensitivity (catching positives) and specificity (avoiding false alarms).

A perfect model would reach the top-left corner of the plot (100% true positives with 0% false positives), while a random guess follows the diagonal line. The area under the curve (AUC) summarizes overall performance in a single number between 0.5 and 1.

ROC curves are especially useful for comparing models and understanding behavior across different operating points without committing to one fixed threshold.

Example

A spam filter model might produce a ROC curve showing that at a low threshold it catches 95% of spam but also flags 20% of legitimate emails as spam, while a higher threshold reduces false positives to 5% but misses more spam.

Why it matters

ROC curves help developers choose and compare models for real-world use cases where the cost of false positives versus false negatives matters, making AI systems more reliable in fields like healthcare and fraud detection.

Frequently asked questions

It means the model performs better than random guessing at distinguishing between the two classes.