Classification stratificationColumn not supported for boolean column

[CESARDELATORRE]

For not balanced datasets, with stratified splitting, the data is divided in such a way that a percentage of each target column value is put in both training and test dataset.

However, the following line of code throws an error if the column 'Label' is Boolean, which is very common for binary classification.

(trainData, testData) = classification.TrainTestSplit(data, testFraction: 0.2, stratificationColumn: "Label");

It would work if the Label column would be float or other types.

I might be missing something, but why is Boolean not supported for the stratificationColumn?
Can we support it since it can be a common scenario for binary classifications?

[Zruty0]

Stratification column is not what you think. Actually, if you read the documentation, it states that

If two examples share the same value of the stratificationColumn (if provided), they are guaranteed to appear in the same subset (train or test). Use this to make sure there is no label leakage from train to the test set

As you can see, it is ALWAYS a bad idea to use Label as a stratification column, and ALWAYS a bad idea to use a Boolean column as a stratification column.

That said, currently we only support float, key and string values for stratification (and I'm not sure about string). We should expand the coverage of Hash estimator (see #1031)

[CESARDELATORRE]

It would be good to be able specify the "stratification column" as a parameter in the TrainTestSplit().
The 'samplingKeyColumn' parameter is almost the opposite concept.

[codemzs]

I agree with Pete. Closing this.