0

I am using the h20 cluster to train the model using tuned random forest and plot pdp plots using the below code

tx = tuned_rf.explain(test_data, top_n_features=5, include_explanations='pdp')

This is returning pdp plots in graphical format the graph is features vs mean response but I want to get the pdp plots in tabular form(below shows the expected output)

Feature          Mean response 
Sub-feature-1      0.18
Sub-feature-2      0.15

but right now i am getting the same in graphical format I want to get the output in the tabular format how can I do this using h20 cluster

rqqa
  • 35
  • 6

1 Answers1

0

have a look at H2O's partial_plot. With plot=False you can get just the data that you want for each feature (and target in case of multinomial classification) separately. The features in explain method are sorted using variable importance that you can retrieve for most models by using the varimp method.

Tomáš Frýda
  • 546
  • 3
  • 8