XGBoost - features_importance() and NULL handling not implemented?
Hello - I have created and fit a model using XGBoost using Vertica 12 CE. I can call methods like .score and .roc_curve, however .features_importance returns an error:
FunctionError: Method 'features_importance' for 'XGBoostClassifier' doesn't exist.
It seems that it should (https://www.vertica.com/python/documentation_last/learn/XGBoostClassifier/index.php) and indeed does work with the Random Forest Classifier.
I believe that XGBoost (the algorithm) handles missing values without needing to impute, however Vertica's implementation appears to reject rows where a predictor (X) contains NULL. Is that correct?