Attributes

tree_count_

Purpose

Return the number of trees in the model.

This number can differ from the value specified in the --iterations training parameter in the following cases:

• The training is stopped by the overfitting detector.
• The --use-best-model training parameter is set to True.

Type

int

feature_importances_

Purpose

Return the calculated feature importances. The output data depends on the type of the model's loss function:

• Non-ranking loss functions — PredictionValuesChange
• Ranking loss functions — LossFunctionChange

If the corresponding feature importance is not calculated the returned value is None.

Use the get_feature_importance function to surely calculate the LossFunctionChange feature importance.

Type

numpy.ndarray

random_seed_

Purpose

The random seed used for training.

Type

int

learning_rate_

Purpose

The learning rate used for training.

Type

float

feature_names_

Purpose

The names of features in the dataset.

Type

list

evals_result_

Purpose

Return the values of metrics calculated during the training.

Type

dict

Output format:

{pool_name: {metric_name_1-1: [value_1, value_2, .., value_N]}, .., {metric_name_1-M: [value_1, value_2, .., value_N]}}

For example:

{'learn': {'Logloss': [0.6720840012056274, 0.6476800666988386, 0.6284055381249782], 'AUC': [1.0, 1.0, 1.0], 'CrossEntropy': [0.6720840012056274, 0.6476800666988386, 0.6284055381249782]}}

best_score_

Purpose

Return the best result for each metric calculated on each validation dataset.

Type

dict

Output format:

{pool_name_1: {metric_1: value,..., metric_N: value}, ..., pool_name_M: {metric_1: value,..., metric_N: value}

For example:

{'validation': {'Logloss': 0.6085537606941837, 'AUC': 0.0}}

best_iteration_

Purpose

Return the identifier of the iteration with the best result of the evaluation metric or loss function on the last validation set.

Type

int or None if the validation dataset is not specified.

classes_

Purpose

Return the names of classes for classification models. An empty list is returned for all other models.

The order of classes in this list corresponds to the order of classes in resulting predictions.

Type

list (an empty list is returned for non-multiclassification models)