Extract feature importances from MLlib GBT model using Dataiku API

Options
AnnaProba
AnnaProba Registered Posts: 9

Hello,

We are using a Spark MLlib GBT model (GBTClassifier) in our project. However, the visual analysis does not provide information about important features of MLlib models. I know that featureImportances are part of the MLlib GBTClassifier and I could extract them myself if I could access the serialized model via Dataiku API in Python or Scala. Is it possible? Where is the serialized model stored?

Setup Info
    Tags
      Help me…