labdmitriy / ml-lab Goto Github PK

Hi @labdmitriy.

Thank you for your excellent job. However, the use of train_size and test_size in the way you've used doesn't make sense. In your code, train_size actually means train_group_size, however, it should be the number of samples in the training set from each group. I appreciate you if you implement that feature too.

Actually, I am working on such a dataset as you can in the image attached. I need to perform cross-validation with a let's say test_size=4 from each group. The code may need to handle the cases for which the split cannot be done so it should ignore them or add all to train (it can be defined by the user, as an argument).

Thank you