What is Anomaly Detection

Definition

Anomaly detection is the identification of data points, events, or patterns that deviate significantly from expected behaviour, using statistical or machine-learning methods to flag rare or unusual observations that may indicate fraud, faults, or other issues worth investigating.

« Back to Glossary Index

Flags unusual events that may signal fraud, faults, or attacks
Catches problems early by surfacing deviations from normal patterns
Reduces manual monitoring through automated detection
Adapts to many domains from security to equipment health

Glossary: Data Observability Tools

Real World Example

A bank uses anomaly detection to flag a transaction far outside a customer's normal pattern, prompting a verification step that catches fraudulent activity the moment it deviates from expected behaviour.

FAQs

How does anomaly detection work?

It learns or defines what normal looks like, then flags observations that deviate significantly from that expected behaviour.

What methods are used for anomaly detection?

Statistical thresholds, clustering, isolation forests, and neural approaches like autoencoders are common.

Where is anomaly detection applied?

Fraud detection, security monitoring, equipment fault detection, and quality and data-quality monitoring.