DriftReport
Models & Types · AiDotNet.DriftDetection
The attributed drift assessment produced by a DriftMonitor over a checked stream.
For Beginners
After training, the model’s errors and predictions on held-out data are streamed through a drift monitor calibrated on the training data. This report says whether that held-out stream already looks like it is drifting away from what the model learned, and — the useful part — whether the drift is in the errors (concept) or the inputs (covariate).
Constructors
| Constructor | Summary |
|---|---|
DriftReport() |
Properties
| Property | Summary |
|---|---|
DriftSource Source { get; set; } |
Gets which shift the detected drift is attributed to. |
bool DriftDetected { get; set; } |
Gets whether any drift (concept or covariate) was detected. |
bool WarningDetected { get; set; } |
Gets whether a warning-level (pre-drift) signal was raised on either lens. |
int FirstDriftIndex { get; set; } |
Gets the index in the checked stream at which drift was first flagged, or -1 when none was. |
int ObservationsChecked { get; set; } |
Gets the number of observations streamed through the monitor during the check. |