I have a question about interpreting the feature importance values obtained using xgboost. Do these values show the local impact of each parameter on the model or they show the global impact? By global impact I mean it considers interactions among parameters or not while these effects are not considered in local impact.
