Anomaly Detection Threshold Tuning

Explore the tradeoff between false positives and false negatives using ROC curve analysis

Detection Threshold

50

Lower threshold = more alerts (higher recall, more FPs) | Higher threshold = fewer alerts (lower recall, fewer FPs)

True Positive Rate

0%

Recall / Sensitivity

False Positive Rate

0%

Type I Error

Precision

0%

TP / (TP + FP)

F1 Score

0.00

Harmonic mean

ROC Curve (Receiver Operating Characteristic)

Confusion Matrix

Predicted Positive
Predicted Negative
Actual Positive
True Positive
0
0%
False Negative
0
0%
Actual Negative
False Positive
0
0%
True Negative
0
0%
Understanding the Tradeoff: The ROC curve shows all possible threshold settings. The current threshold is marked in red. Moving left on the curve (lower threshold) catches more threats but increases false alarms. Moving right (higher threshold) reduces noise but misses real threats. Ideal threshold depends on your operational context: high-security environments tolerate more FPs, while high-volume SOCs need fewer alerts.