Fig. 5.

From [31]; Evaluation (total train and test) time for XGBoost, CatBoost, and LightGBM as number of features increases, average of 100 random customers; shows improvement in CatBoost false positive rate when all summary statistics are employed