Global feature importance in XGBoost R using SHAP values

Question

I would like to know if there is a method to compute global feature importance in R package of XGBoost using SHAP values instead of GAIN like Python package of SHAP. I mean, in XGBoost for Python there is a function to compute SHAP values at global level making the mean absolute of the SHAP value for each feature.

Answer 1

You can do:

shap_contrib <- predict(model,newdata,predcontrib=T)

in order to get the SHAP values directly from XGBoost .

Once you get that, it's just a matter of doing:

mean_shap <- colMeans(abs(shap_contrib))