Ensemble in h2o (random forest) -multinomial distribution

0
votes

I have a problem with building an ensemble using h2o and Random Forest as one of the components. I estimate the models as in h2o's documentation with the only difference being the dependent variable being a multiclass variable. When I try to build an ensemble I get the following error:

Error: water.exceptions.H2OIllegalArgumentException: Don't know how to set the distribution for a multinomial Random Forest classifier.

I tried googling it but with no success. Anyone knows what to do in such a case?

1
rmachine-learningrandom-foresth2oensemble-learning
Please provide a sample of your code and your data – desertnaut
what version of h2o r are you using? you may need to upgrade to the latest stable release of H2O to get support for multiclass. example code here docs.h2o.ai/h2o/latest-stable/h2o-docs/data-science/… – Lauren
That helped, thanks! – abu

1 Answers

1
votes

Multiclass support for Stacked Ensemble was only released very recently (in H2O 3.16.0.1), so you'll need to upgrade your H2O package for this to work.

You can use update.packages("h2o") to upgrade to the latest CRAN version.