Assessments¶
Assessments are the primary workspaces within Falkonry Patterns for applying models, organizing signals, defining events, and interpreting system behavior. While they are tightly linked to a parent Datastream, Assessments provide a focused environment to address specific analytical questions without interfering with other work in the same dataset.
Each Assessment allows users to:
- Select and organize relevant signals
- Assign signal properties (e.g., thresholds, gap handling, data rate adjustments)
- Build and evaluate machine learning models
- Create and label events
- Monitor model outputs (predictions, confidences, explanations)
- Visualize and analyze model performance
Purpose and Role¶
An Assessment is designed as a subset of a Datastream and serves as an isolated workspace to explore, model, and monitor a particular use case. For example, you might create:
- An Assessment for detecting early failure conditions
- A separate one for tracking product quality trends
This structure allows for experimentation and iteration on specific modeling tasks without affecting other workflows.
Assessment Components¶
| Component | Description |
|---|---|
| Signals | Each Assessment defines a set of signals selected from the parent Datastream. These may include signal properties specific to the Assessment, such as thresholds, sendOnChange, or smoothing parameters. |
| Signal Groups | Signal Groups Glossary |
| Timeline | Visualizes signal data, predictions, confidences, and labeled events across time. |
| Models | Models Glossary |
| Output Signals | Models produce three types of outputs: - Predictions: Categorical condition values over time - Confidence Scores: Numeric signal (0.0–1.0) indicating certainty - Explanation Scores: Numeric signal (-1.0 to +1.0) showing signal influence |
| Events (Facts) | Events Glossary |
| Event Groups | Event Groups Glossary |
| Evaluations | Evaluation Glossary |
| Live Model History | Shows the timeline and details of all models that have been used for live monitoring within the Assessment. |
| Entity Display | Provides tabs for individual entities, showing only the events, model outputs, and signal data relevant to the selected entity. |
Workflow & Management
- When a new Datastream is created, a default Assessment (Assessment 1) is automatically added.
- Users can create additional Assessments, name them for clarity (e.g., Bearing Failure Monitoring), and configure their signal sets.
- Models, events, evaluations, and outputs live within their respective Assessments.
- All outputs can be accessed via API for integration with workflows, rules engines, reporting systems, external dashboards, and outbound MQTT connections.
Model Outputs from Assessments are stored as signals, enabling advanced use cases such as chaining models or using condition outputs as inputs to other models — a "model-of-models" design pattern.
Note
Rules, Insights, and Calculations Transform models are all scoped to individual Assessments within a specific Datastream. This structure is important to understand when working with the Falkonry APIs, as model operations and outputs are tied to their respective Datastream and Assessment contexts.