Including Cook's D obtained from GWR leads to an enhancement in the accuracy of XGBoost,
I have the following scenario:
1. Initially, I input the 5 features and target variable into XGBoost, resulting in a calculated R2 of 0.8.
2. Next, I input the same 5 features and target variable into GWR, where Cook's D is computed for each observation.
3. Subsequently, I include the 5 features, target variables, and Cook's D into XGBoost, yielding a calculated R2 of 0.9.
What contributes to the improvement in XGBoost accuracy when Cook's D is integrated as a feature?
------------------------------
jamal numan
------------------------------