The scenario
A production predictive maintenance system must monitor model inputs, predictions, and drift to ensure reliability and trigger alerts.
What is in this drawing
Read the decisions behind it.
01
Separate Prediction and ActualOutcome tables to compare forecasts with observed equipment failures
02
Model feature distributions and drift metrics as first-class entities to support automated monitoring
03
Link DataQualityIssue to Reading to track upstream data quality problems
Use when designing a schema for ML observability, drift detection, or model performance tracking in industrial IoT or maintenance domains.
More like this