Registered users can ask their own questions, contribute to discussions, and be part of the Community!
In the following article https://explained.ai/rf-importance/, it is stated that the Feature Importance of Random Forests can be biased towards continuous variables. If I understand correctly DataIKU uses the standard scikit-learn FI, also when other models are used as explained here: https://community.dataiku.com/t5/Using-Dataiku-DSS/Variables-importance/m-p/1962
What is your opinion on this article and if you agree that the FI is biased, do you have a solution within DataIKU?
Post a Question