Checking the Training Data

The Data → Check Training Data menu option allows checking whether the training data used are distributed in a space-filling way. The windows "Check Training Data: Inputs" and "Check Training Data: Outputs" are opened. In both windows, the measured values of certain inputs/outputs are plotted over other inputs/outputs.

Proceed as follows.

In the ISP view, select Data → Check Training Data.

The "Check Training Data: Inputs" and "Check Training Data: Outputs" windows open.
In the plots of the "Check Training Data: *" windows, search for outliers.
If you found one or more outliers, click in the respective plot and draw a rectangle around the points.

The selected points are highlighted in color in all plots.
To mark the points within a specific rectangle as outliers, right-click the edge of that rectangle and select Mark from the context menu.

Mark in the "All Outputs" area marks the outlier for all outputs.

Mark in an "<output_name>" area marks the points in the rectangle only for the respective output.

Mark Outside marks all points outside the rectangle.
Use Remove Rectangle to remove an existing rectangle.

Marked points remain marked.
Select Extras → Delete Marked Points / Retrain to delete the marked points and retrain the model.