Receiver Operating Characteristic (ROC) AUC

INFO

Represents the area under the receiver operating characteristic (ROC) curve, assessing classification performance.

How It Works

• Plots the True Positive Rate (TPR) against the False Positive Rate (FPR) across thresholds
• $TPR=\frac{\text{True Positive}}{(\text{True Positive}+\text{False Negative})}$
• $TPR=\frac{\text{False Positive}}{(\text{False Positive}+\text{True Negative})}$
• AUC quantifies the model’s ability to distinguish between classes

What to Look For

• AUC close to 1 = excellent performance
• Threshold-independent
• Best for binary classification tasks

Application Models

• Convolutional Neural Network (CNN)
• Deep Neural Network (DNN)
• Transformer