Table 5: List of model hyperparameters
Hyperparameter,Description,Value
objective,Type of learning task to be carried out.,“reg:linear”
tree_method,Algorithm used for constructing trees.,“exact”
base_score,Initial model bias.,mean of house price in training data
max_depth,Maximum tree depth.,16
learning_rate,Shrinkage applied to tree weights.,0.05
n_estimators,Number of boosting rounds.,400
gamma,Minimum loss required to make a further partition in the tree.,0
subsample,Number of rows to randomly sample (without replacement when training a single tree).,0.98
reg_lambda," regularisation term on tree weights.",2
reg_alpha," regularisation term on tree weights.",0.1
col_samplebylevel,Subsample ratio of columns for a single tree,0.6
col_samplebylevel,Subsample ratio of columns for a single level,0.6
col_samplebylevel,Subsample ratio of columns for a single node,0.6
random_state,Random number seed – ensures reproducibility.,123
Source: ,Office for National Statistics