Back to glossary

ROC Curve

Receiver Operating Characteristic curve; a graph showing the performance of a classification model at all classification thresholds.

AbbreviationROC
Categoryai

Overview

The ROC curve plots the True Positive Rate (Recall) against the False Positive Rate as the decision threshold is varied from 0 to 1.

Key Concepts

  • True Positive Rate (TPR): Same as Recall.
  • False Positive Rate (FPR): FP / (FP + TN).
  • Threshold: The probability cutoff used to decide if a prediction is positive or negative.

Interpretation

A model that makes random guesses will result in a diagonal line. A perfect model will have a curve that hugs the top-left corner of the graph.

Related Terms

AUC (Area Under Curve)RecallConfusion MatrixPrecision-Recall Curve