| ID | product.processing.stream_processor.anomaly_detector |
|---|---|
| Description | Detects anomalies and emits alerts. |
| Key | anomaly_detector |
| Type | component |
| Team | Stream Processing |
| Owner | stream-processing@iotgateway.example |
| Status | active |
| Technologies | Java, Python, Apache Flink |
| Attributes | detection_window: 5m sliding, latency: <1s |
| Tags |
Table of contents
Guide
Anomaly Detector
The Anomaly Detector is the last stage of the streaming pipeline. It takes windowed aggregates from the Window Aggregator, scores them against per-device baselines, and emits alerts when a device drifts outside expected bounds.
Tuned for signal, not noise
Thresholds are per-device and adaptive. A cold-storage sensor and a rooftop sensor have very different “normal” — a single global threshold would either miss real faults or drown operators in false positives.