Table III.

Description of the Extreme Gradient Boosting algorithm hyperparameters and ranges.

Hyperparameter	Definition	Significance/usefulness	Grid search range
ETA	ETA is the shrinkage of the learning rate at each step	ETA prevents the model from overfitting by scaling contribution of each tree.	From 0.01 to 1 at a step size of 0.01
Max Depth	Maximum depth of each tree	Max depth affects the complexity of each tree and the overfitting of the model. Deeper trees can overfit the data.	1, 2, 3, 4 and 5
Minimum Child Weight	The minimum weight required in order to create a new node in each tree	Adjusting the value can prevent overfitting and reduce the model complexity.	1, 2, 3, 4 and 5
γ	Minimum loss reduction required to create a further partition on a tree's leaf node	Increasing γ causes the model to be more conservative.	1, 2, 3, 4 and 5
Nround	The number of training rounds within the model	Increasing Nrounds can reduce biases and variance in a model.	Varies depending on model error/loss and 200 Nround early stopping

ETA, estimated time of arrival (also known as the learning rate in R user documentation for XGBoost).