I'd like to have the ability for the minimum split size for tree models to be based upon a % of the training sample size vs a integer number. So for example I could ask that the minimum split size would be 10% of the training samples size, rounded up.