weights_column and unbalanced data - h2o.glm

275 views

Skip to first unread message

Gabriela de Queiroz

unread,

Sep 28, 2015, 5:33:06 PM9/28/15

to H2O Open Source Scalable Machine Learning - h2ostream

My data is very unbalanced. I've been using RF and the balance_classes option. I'd like to use the same idea to perform glm (logistic regression).

I know that there is an argument weights_column on h2o.glm, but I don't have this column on my data.

Any advice on how to deal with unbalanced data when running h2o.glm?

Thanks,

Avni Wadhwa

unread,

Oct 1, 2015, 7:16:51 PM10/1/15

to H2O Open Source Scalable Machine Learning - h2ostream

Hi,

The following two examples will help you better understand observation weights in glm. Let me know if you have questions after looking at the examples.

https://github.com/h2oai/h2o-3/blob/f373341d3881cb3168a3c154a5bc50f22577f98d/h2o-r/tests/testdir_algos/glm/runit_GLM_weights1.R

https://github.com/h2oai/h2o-3/blob/f373341d3881cb3168a3c154a5bc50f22577f98d/h2o-r/tests/testdir_algos/glm/runit_GLM_weights2.R

Thanks,

Avni

Reply all

Reply to author

Forward

0 new messages