Skip to main content

Anomaly Detection System

Implement automated anomaly detection for metrics and KPIs using statistical methods, ML models, alerting, and root cause analysis.

Complexity: Complex | Duration: 30m+ | Category: Devops

Tags: analytics anomaly-detection monitoring ml

Workflow Diagram​

Steps​

Step 1: Baseline creation​

Agent: data

scientist - Calculate normal ranges from historical data

Step 2: Method selection​

Agent: data

scientist - Choose statistical (Z-score, IQR) or ML (Isolation Forest, Autoencoders)

Step 3: Model training​

Agent: data

scientist - Train anomaly detection model

Step 4: Real​

Agent: time detection

data-engineer - Apply model to incoming data

Step 5: Anomaly scoring​

Agent: data

scientist - Score anomalies by severity

Step 6: Alerting​

Agent: data

engineer - Alert on high-severity anomalies

Step 7: Root cause analysis​

Agent: data

analyst - Investigate potential causes

Step 8: Feedback loop​

Agent: data

scientist - Incorporate feedback to reduce false positives

Usage​

To execute this workflow:

/workflow devops/anomaly-detection-system.workflow

See other workflows in this category for related automation patterns.