4

I have a data set with over 400 features that I am estimating with GBM using H2O atop R. When I use the variable importance function (h2o.varimp) it only shows me the head and tail of the full ranked variable list. Is there a way to have the entire list displayed?

dj_ski_mask
  • 55
  • 1
  • 5

2 Answers2

9

This is not specific to variable importance, this is just how H2O displays H2O Frames in the R console. If you want to view the whole frame, you could convert it to an R data.frame and then print it. 

df <- as.data.frame(h2o.varimp(model))
print(df)
Erin LeDell
  • 8,704
  • 1
  • 19
  • 35
0

(summary) will show all stats with extraction from h2o.varimp. Then save the table of variable importance

mymodel <- summary(model)

write.table(mymodel, file = "mymodel.txt", sep = "\t", quote = FALSE, row.names = TRUE)
tylers
  • 1
  • 2
  • Thanks. I am curious, is that command also how you would permanently store a model object for scoring new data (in H2O format) in the future? – dj_ski_mask Apr 24 '17 at 15:20
  • 1
    @dj_ski_mask Check out `h2o.saveModel()` and `h2o.loadModel()` for binary models. Those are the easiest to use, however for productionizing models, check out `h2o.download_pojo()` or `h2o.download_mojo()`. http://docs.h2o.ai/h2o/latest-stable/h2o-docs/pojo-quick-start.html – Erin LeDell Apr 24 '17 at 21:30