How to apply undersampling data balancing, setting a parameter?

Question

Im using H2O library and i want to apply undersampling data balancing, setting parameter 0.8. How can i do this? I wrote this command:

from h2o.estimators.gbm import H2OGradientBoostingEstimator
cov_gbm = H2OGradientBoostingEstimator(balance_classes = True)

but balance_classes = True use a random parameter. I want to make this parameter 0.8.

I will appreciate any help. Thanks.

Answer 1

You need to provide one more argument: class_sampling_factors . Assuming there are 2 classes and You want to undersample the first class then use:

cov_gbm = H2OGradientBoostingEstimator(balance_classes = True, class_sampling_factors = [0.8, 1])

class_sampling_factors must be a list of floats. Each element of list determines factor of sampling per class.