Churn prediction, which values lead to the decision?

[sabsab]

Churn prediction works very well with AutoML as in the Hub example (Hub Churn example).
Is it also possible to name or give a value to those elements/columns that are particularly crucial for the classification (churn 1 or 0)?
For the analyst it would be interesting to know which values should be influenced best.

Many greetings
Bernd

[victor_palacios]

Hello,
The article on explainable machine learning (XAI) is here. This will show you the different techniques to see which columns are important for your prediction.

[sabsab]

That’s cool! I’ll have to learn the basics, but that’s what it should be.

[Mpattadkal]

With KNIME 4.6, we have the new XGBoost Integration, it performs classification and regression and provides the feature importance table as well to the user

For example, if you use the below-given node for classification, the second output port will explain the feature importance

XGBoost Tree Ensemble Learner

[sabsab]

Hi @Mpattadkal,
thank you very much, good to know.
Unfortunately Knime crashed on me when I upgraded to 46, now back at 452. Still need to address the problem.