Model studio - Use a variable to perform partitions in the train and test sets

[dcortell]

Hi all

I've imported a dataset in model studio to build a pipeline on that. Within the dataset, I have a nominal variable which I want use to stratify both the train and test set with specific proportions (example: 30% class 1, 50% class2, 20% class3).

Now, I'm not able to find documentation explaining how to perform that stratification nor in the Data tab nor in the pipeline.

There is the chance to define the variable with role as "Partition": but then it only allow to assign one class of the variable to one specific set only (ex. class1 to its totality to the training set..)

Does anyone know if there is a way to setting a specific stratification in both the sets using a variable from the dataset?

[dcortell · Accepted solution]

Checked, but it seems with current edition of Model studio I can't choose multiple class variable levels to map to a single partition level (training/validation/test) from with-in Model Studio (VML), so I will have do this before the dataset import.