Overview
Most monitors in Monte Carlo support automated thresholds, powered by machine learning and artificial intelligence.
For any automated threshold, there are 4 components of the system. Two of these components are managed exclusively by Monte Carlo, and two can be adjusted by users to tune thresholds. In most cases, the default settings are sufficient and no adjustments to any settings are needed.
Components of automated thresholds
| Component | Description | User configurable? |
|---|---|---|
| Models | Behind each monitor is an ensemble of anomaly detection models, optimized for different cases and profiles of data. | Managed by Monte Carlo |
| Model selection | To produce an accurate threshold, the best suited model is chosen based on patterns observed in that series of data. | Managed by Monte Carlo |
| Training data | Models are frequently retrained using a rolling window of recent data. Anomalies and periods of unusual behavior can be excluded to ensure thresholds are not unnecessarily widened. Learn more. | Yes |
| Sensitivity | Sensitivity of a threshold can be adjusted higher or lower. Low sensitivity will widen the threshold, resulting in fewer alerts. High sensitivity will tighten the threshold, resulting in more alerts. Learn more. | Yes |
Updated 1 day ago