Baseline and production data are required to monitor model drift.
Upload baseline data for a model
Baseline data are required to monitor drift. Follow these steps to add or replace
baseline data for an existing model. For information on baseline data requirements,
go to Requirements for baseline data.
- From the Deployments page, select the deployment to open the tabs on the deployment area page.
- Open the Models tab.
- Open the settings of the model
. - Select Upload baseline data or Replace baseline data.
- Browse to the file that contains the baseline data and specify which columns to upload for the predicted response and predictors.
- Select Upload.
Note
You can also upload baseline data for a model from the model repository.
Baseline data status indicators
From the Models tab in the deployment area, you can see the list of your models and whether they have valid baseline data.
| Icon | Details |
|---|---|
 |
The baseline data upload is valid and you have at least one row of missing or nonmissing values in your baseline data for all predictors in the model. |
 |
The baseline data upload is valid but you are missing data for some of the predictors in the model. |
Check the System Audit
Log
area from the left pane navigation to see specific
baseline data errors.
Send production data
To score a model, Minitab Model
Ops requires new production data. You can create a
Minitab Connect table that contains the new predictions, or you can send the data
via Python code, a terminal, or a script in a terminal. For help on generating the
required code format for prediction requests, see the sample code from the
Integration tab.For
information on production data requirements, go to Requirements for production data.
Note
Prediction requests with prior date/time values beyond three months are not logged; thus, are not available in the drift and stability reports.
