AI Correlation

Keep’s AI correlation engine provides a distinctive approach to fully AI-driven alert correlation. By using historical alert data as its training dataset, the system intelligently classifies new alerts and assigns them to appropriate incidents.

The AI correlator runs on cycles, each iteration cycle completes in 5-15 minutes:

1. Model trained based on historical data.
2. Model is evaluated.
3. All unassigned alerts are clustered and added to incidents when their confidence score exceeds the threshold.

Configuration UI:

Incident with alerts correlated by AI:

Check the demo on a playground: https://playground.keephq.dev/ai

To activate the feature for your on-premises tenant, please talk to us.

​

Frequent questions:

Model used: proprietary model developed and hosted by Keep.
Training dataset: tenant’s alerts and incidents.
Privacy: tenant’s data is used only for training of the model for the same tenant. Data is not mixed between tenants for training.