1

I want to know the variable importance of our models, which has been built on 3 models (random forest, deep learning and gradient boosted tree). I know you should use "weight by tree importance" to asses the variable importance for just the random forest. However, how do you assess the variable importance for the three models combined?

thanks in advance!

Frank Li
  • 5
  • 3

1 Answers1

1

you can use Explain Predictions. It calculates a global feature importance scores, which are the ones displayed in AutoModel. It also excepts ensembles like Stacking or Voting.

Best, Martin

Martin Schmitz
  • 141
  • 2