Golden Features are new features constructed from original data which have great predictive power. Please see the Golden Features section in the documentation for more details about how are they constructed.
The Golden Features are constructed only once during AutoML fit. They are saved in results_path in golden_features.json file.
After creation of the Golden Features they are added to the original data and following algorithms are trained:
Xgboost
CatBoost
LightGBM.