Survey banner
Share your feedback on the Dataiku documentation with this 5 min survey. Thanks! TAKE THE SURVEY

Enable only certain features for MES drift model and drift metrics

cmjurs
Level 3
Enable only certain features for MES drift model and drift metrics

Using an evaluate recipe, I'd like to be able to turn off certain features during the calculation of the drift model and the univariate drift metrics. Im seeing a section in "advanced" which allows for some custom variables. Maybe there is a setting for this were I can apply a list of enabled features??

 

2023-10-02 14_43_09-(9) mes feature tweak _ Jupyter notebook _ Dataiku — Mozilla Firefox.png


Operating system used: Ubuntu


Operating system used: Ubuntu

0 Kudos
3 Replies
fsergot
Dataiker

Hello @cmjurs ,

You can indeed exclude a feature from the data drift computation in the Evaluate recipe: in the advanced section, you need to check the "Feature handling" box and you can then see all the evaluation dataset columns.

 

 

For each, you have the choice of:

  • Chec/Uncheck => Inlcude or Exclude the column from being considered in datadrift
  • When checked, you can choose the behaviour:
    • Auto => let Dataiku decides if the feature is to be taken into account and as which data type (default). Basically, DSS will consider all the model features (and ignore columns that are not model features)
    • Numerical => Force consider the feature as a numerical
    • Categorical => Force consider the feature as a numerical

2023-10-03 09_09_24-(1) evaluate_on_pokemon_for_scoring_scored - Recipe _ Dataiku.png

 

 

 

0 Kudos
cmjurs
Level 3
Author

Hi @fsergot 

Thanks so much for your answer. I'm glad to see that there is something in the UI to handle these situations. Unfortunately in my advanced options, those check boxes are not there...?

evaluate-adv-opts.png

 

Im on version 11.4.4.

0 Kudos
fsergot
Dataiker

You pointed the right issue: the capability to customize the feature handling and selection was added in 12.0.

0 Kudos