select features & xgboost with Shap

[gdupret]

Use Shap to evaluate the feature importance in XGBoost models

Shap is not available from the sklearn API to XGBoost so I added a wrapper around XGBoost to include it

Select the features to be tested

Sometimes some features are known to be important, so there is no benefit in testing them. Removing these features allows to save computing time and memory.

Miscellaneous

rewrite np.int to int, np.float to float and np.bool to bool to avoid deprecation warnings