Hyperparameters

.. Parameter documentation master file.

Parameters

Specifications of the C-GB parameters.

Base model

• n_estimators : int, default = 100

  • Number of Decision Regressor Tree to build an ensemble.

• subsample : float, default = 1.0

  • The division of samples for fitting the base learners.

• max_features : {‘auto’, ‘sqrt’, ‘log2’}, int or float, default= None

  • The number of points for splitting the tree.

    • auto , sqrt >> sqrt(n_features)
    • log2 >> log2(n_features)
    • None >> n_features

cgb_clf

• loss: {log_loss, ls}, default = log_loss

  • The loss function for optimization. For the Multi-class/Binary classification, it should be deviance.

cgb_reg

• loss: {log_loss, ls}, default = ls

  • The loss function for optimization. For the regression it should set to ls.

• metric : {rmse, r2_score}, default = rmse

  • It returns the error of the model. rmse will return the average R2 score.