Machine learning for anomaly detection (demo with BGP prefixes received)

HealthBot and machine learning

HealthBot supports machine learnings for anomaly detection and for outlier detection.

HealthBot supports the following machine learning algorithms for anomaly detection:

• Three-sigma rule
• k-means for anomaly detection

HealthBot supports the following machine learning algorithms for outlier detection:

• DBSCAN (Density-Based Spatial Clustering of Applications with Noise)
• K-fold Three-sigma ("K-Fold Cross-Validation" using "Three-sigma")

Anomaly detection and outlier detection are both about detecting anomalies.
In HealthBot terminology:

• anomaly detection is time based. It compares new data points from a device vs data points collected from the same device during a learning period.
• outlier detection is group based. It analyzes data from a device during a learning Period vs data from other devices during the same learning period

Machine learning demo

Please refer to this other repository to see:

• machine learning 101
• machine learning demo with HealthBot and Junos