TabularPredictor.simulation_artifact

TabularPredictor.simulation_artifact(test_data: DataFrame = None) → dict

[Advanced] Computes and returns the necessary information to perform zeroshot HPO simulation. For a usage example, refer to https://github.com/autogluon/tabrepo/blob/main/examples/run_quickstart_from_scratch.py

Parameters:

test_data (pd.DataFrame, default = None) – The test data to predict with. If None, the keys pred_proba_dict_test and y_test will not be present in the output.

Returns:

simulation_dict – The dictionary of information required for zeroshot HPO simulation. Keys are as follows:

pred_proba_dict_val: Dictionary of model name to prediction probabilities (or predictions if regression) on the validation data pred_proba_dict_test: Dictionary of model name to prediction probabilities (or predictions if regression) on the test data y_val: Pandas Series of ground truth labels for the validation data (internal representation) y_test: Pandas Series of ground truth labels for the test data (internal representation) eval_metric: The string name of the evaluation metric (obtained via predictor.eval_metric.name) problem_type: The problem type (obtained via predictor.problem_type) problem_type_transform: The transformed (internal) problem type (obtained via predictor._learner.label_cleaner.problem_type_transform,) ordered_class_labels: The original class labels (predictor._learner.label_cleaner.ordered_class_labels) ordered_Class_labels_transformed: The transformed (internal) class labels (predictor._learner.label_cleaner.ordered_class_labels_transformed) num_classes: The number of internal classes (self._learner.label_cleaner.num_classes) label: The label column name (predictor.label)

Return type:

dict