Building XGBoost models

In the first two sections, you learned how XGBoost works under the hood with parameter derivations, regularization, speed enhancements, and new features such as the missing parameter to compensate for null values.

In this book, we primarily build XGBoost models with scikit-learn. The scikit-learn XGBoost wrapper was released in 2019. Before full immersion with scikit-learn, building XGBoost models required a steeper learning curve. Converting NumPy arrays to dmatrices, for instance, was mandatory to take advantage of the XGBoost framework.

In scikit-learn, however, these conversions happen behind the scenes. Building XGBoost models in scikit-learn is very similar to building other machine learning models in scikit-learn, as you have experienced throughout this book. All standard scikit-learn methods, such as .fit, and .predict, are available, in addition to essential tools such as train_test_split, cross_val_score, GridSearchCV, and RandomizedSearchCV...