Skip to content

Falkonry Time Series Intelligence (TSI)

Evaluate Model

Falkonry Time Series Intelligence (TSI)

Home
Introduction
Introduction
- Motivation
- Architecture
User Guide
User Guide
- Introduction
- Administration
  Administration
  - Account Management
  - User Management
  - API Access Maangement
  - Private Storage
  - Connections
    Connections
    
    Overview
    
    Inbound Connections
    
    Outbound Connections
- Signals Management
  Signals Management
- Reports
  Reports
  - Overview
  - Organize and Locate Reports
  - Insights Reports
  - Sharing Reports
  - Report Widget Types
    Report Widget Types
    
    Overview
    
    Notes Widget
    
    Timeline Widget
    
    Signal Value Distributions (SVD) Widget
- Rules & Alerts
  Rules & Alerts
- Calculations Transform
  Calculations Transform
  - Overview
- Insights Anomaly Detection
  Insights Anomaly Detection
- Patterns Classification
  Patterns Classification
Tutorials
Tutorials
APIs Reference
APIs Reference
- Overview
- Administration
  Administration
- Signals Management
  Signals Management
- Reports
  Reports
  - Deep linking
- Rules & Alerts
- Calculations Transform
- Patterns Classification
  Patterns Classification
  - Entities
  - Live Monitoring
- Integrations
  Integrations
  - Alert Notification via email using Zapier Integration
Resources
Resources
- Glossary
  Glossary
  - Time Series
  - Falkonry TSI Terms
- Best Practices
- Chart Gestures
- Search using Regex
- User Forum

Model Evaluation (via API)¶

After the learning process, you can evaluate the model's performance.

Use the ANOMALYEVAL API flow to evaluate the trained model over specified time ranges
- A unique model identifier must be specified to indicate which Insights model is to be evaluated.
- Once the evaluation completes successfully, the results are published to the corresponding model’s /anomaly score signal for downstream use.
The evaluation process produces an "Anomaly Score" signal for each monitored signal, which represents the degree of deviation from the learned normal behavior. Scores above 3 generally indicate noteworthy or rare behaviors.

How to evaluate an anomaly model using API has more details.