I have used h2o package for GBM and it has worked brilliantly, but I am a bit lost when it comes to marginal contribution calculation using h2o.partialPlot.
I wonder why we have to specify the number of bins(default is 20). What if even 100 bins fail to catch the changes at the correct points?
Is there a way where the h2o.partialPlot function itself can figure out the breakpoints?
How does it deal with the NA's/ special values(should be a category) in a continuous variable?