Battle of the Boosting Algorithms

We're going to let XGBoost, LightGBM and Catboost battle it out in 3 rounds: In each round here are the steps we'll follow: You can find the accompanying code here. Let's start with the top findings. LightGBM is the clear winner in terms of both training and prediction times, with CatBoost trailing behind very slightly. XGBoost took substantially more time to train but had reasonable prediction times. We also want to know why the model is making its predictions.