Exploring XGBoost hyperparameters
Extreme Gradient Boosting (XGBoost) is also a tree-based model that is built using a collection of decision trees, similar to a Random Forest. It can also be utilized for both classification and regression tasks. The difference between XGBoost and Random Forest is in how they perform the ensemble. Unlike Random Forest, which uses the bagging ensemble method, XGBoost utilizes another ensemble method called boosting.
Boosting is an ensemble algorithm whose goal is to achieve higher performance through a sequence of individually weak models by overcoming the weaknesses of the predecessor models (see Figure 11.1). It is not a specific model; it’s just a generic ensemble algorithm. The definition of weakness may vary across different types of boosting ensemble implementation. In XGBoost, it is defined based on the error of the gradient from the previous decision tree model. Take a look at the following diagram:
Figure...