2
$\begingroup$

I would like to present the impact of each variable that I use in a binary:logistic model in xgboost. Is there a way or a function in R to know such a thing?

$\endgroup$
1
  • $\begingroup$ You can use this library to help quantify and visualize the impact of each feature in your XGBoost model: github.com/slundberg/shap $\endgroup$
    – Alex Cui
    Mar 28, 2019 at 7:18

1 Answer 1

1
$\begingroup$

This is a good question, because model interpretation is more important than the model itself.

First, you can try to using gblinear booster in xgboost, it's feature importance identical the coefficient of linear model, so you can get some impact direction of each variable.

Second, you can try the monotone_constraints parameters in xgboost, and give some variable the monotic constrain, then compare the result difference.

For more information about monotone_constrains, you can visit this site:https://xgboost.readthedocs.io/en/latest/tutorials/index.html

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.