actableai.classification package

Subpackages

• actableai.classification.models package
  • Submodules
  • actableai.classification.models.tabpfn module
  • Module contents
• actableai.classification.tests package
  • Submodules
  • actableai.classification.tests.test_cross_validation module
  • actableai.classification.tests.test_roc_curve_cross_validation module
  • actableai.classification.tests.test_utils module
  • Module contents

Submodules

actableai.classification.config module

actableai.classification.cross_validation module

class actableai.classification.cross_validation.AverageEnsembleClassifier(predictors)

Bases: object

predict(X) → pandas.core.series.Series

Predicts the class for each sample in X. :param X: DataFrame with features.

Returns

Predicted class for each sample.

Return type

pd.Series

predict_proba(X, *args, **kwargs)

Predict probabilities for each predictor for each class for each sample.

Parameters

X – DataFrame with features.

Returns

List of probabilities for each predictor for each class

Return type

List[np.ndarray]

unpersist_models()

Unpersists all models in the ensemble.

actableai.classification.cross_validation.run_cross_validation(classification_train_task: actableai.tasks.classification._AAIClassificationTrainTask, problem_type: str, explain_samples: bool, positive_label: Optional[str], presets: str, hyperparameters: dict, model_directory: str, target: str, features: list, run_model: bool, df_train: pandas.core.frame.DataFrame, df_test: pandas.core.frame.DataFrame, kfolds: int, cross_validation_max_concurrency: int, drop_duplicates: bool, run_debiasing: bool, biased_groups: list, debiased_features: list, residuals_hyperparameters: Optional[dict], num_gpus: int, eval_metric: str, time_limit: Optional[int], drop_unique: bool, drop_useless_features: bool, feature_prune: bool, feature_prune_time_limit: Optional[float], tabpfn_model_directory: Optional[str], num_trials: int, infer_limit: float, infer_limit_batch_size: int) → Tuple[actableai.classification.cross_validation.AverageEnsembleClassifier, list, dict, List[pandas.core.frame.DataFrame], pandas.core.frame.DataFrame, pandas.core.frame.DataFrame, pandas.core.frame.DataFrame]

Runs a cross validation for a classification task.

Parameters

• classification_train_task – The classification task to run cross validation on.
• problem_type (str) – The problem type. Can be either ‘binary’ or ‘multiclass’.
• explain_samples (bool) – Explaining the samples.
• positive_label (str) – The positive label. Only used if problem_type is ‘binary’.
• presets (dict) – The presets to use for AutoGluon. See https://auto.gluon.ai/stable/api/autogluon.task.html#autogluon.tabular.TabularPredictor.fit for more information.
• hyperparameters – The hyperparameters to use for AutoGluon. See https://auto.gluon.ai/stable/api/autogluon.task.html#autogluon.tabular.TabularPredictor.fit for more information.
• model_directory – The directory to store the models.
• target – The target column.
• features – The features columns used for training/prediction.
• run_model – If True, classification models run predictions on unseen values.
• df_train – The input dataframe.
• df_test – Testing data.
• kfolds – The number of folds to use for cross validation.
• cross_validation_max_concurrency – The maximum number of concurrent processes to use for cross validation.
• drop_duplicates – Whether to drop duplicates.
• run_debiasing – Whether to run debiasing.
• biased_groups – The groups introducing bias.
• debiased_features – The features to debias.
• residuals_hyperparameters – The hyperparameters to use for the debiasing model.
• num_gpus (int) – The number of GPUs to use.
• eval_metric – Metric to be optimized for.
• feature_prune – Whether to prune features.
• feature_prune_time_limit – The time limit for feature pruning. (in seconds)
• tabpfn_model_directory – TabPFN Model Directory.
• num_trials – The number of trials for hyperparameter optimization
• infer_limit – The time in seconds to predict 1 row of data. For example, infer_limit=0.05 means 50 ms per row of data, or 20 rows / second throughput.
• infer_limit_batch_size – The amount of rows passed at once to be predicted when calculating per-row speed. This is very important because infer_limit_batch_size=1 (online-inference) is highly suboptimal as various operations have a fixed cost overhead regardless of data size. If you can pass your test data in bulk, you should specify infer_limit_batch_size=10000. Must be an integer greater than 0.

Returns

Result of the cross validation.

• AverageEnsembleClassifier: The average ensemble classifier.
• list: The feature importances.
• dict: The evaluation metrics.
• list: Probabilities of the predicted classes.
• pd.DataFrame: The training dataframe.
• pd.DataFrame: The test dataframe.

Return type

Tuple

actableai.classification.model module

class actableai.classification.model.ClassificationInference(model: actableai.models.inference.ModelType)

Bases: actableai.models.autogluon.base.AAIAutogluonTabularInference[actableai.classification.model.ClassificationModel, actableai.classification.model.ClassificationMetadata]

class actableai.classification.model.ClassificationMetadata(*, features: List[str], feature_parameters: Dict[str, Any], problem_type: Literal['binary', 'multiclass'], prediction_target: str, is_explainer_available: bool, intervened_column: Optional[str] = None, discrete_treatment: Optional[str] = None, class_labels: List[str])

Bases: actableai.models.autogluon.base.AAIAutogluonTabularMetadata

class_labels: List[str]

problem_type: Literal['binary', 'multiclass']

class actableai.classification.model.ClassificationModel(autogluon_model, df_training: pandas.core.frame.DataFrame, explanation_model=None, intervention_model=None)

Bases: actableai.models.autogluon.base.AAIAutogluonTabularModel

predict_from_proba(df_proba: pandas.core.frame.DataFrame) → pandas.core.frame.DataFrame

set_positive_label_index(positive_label_index: int) → None

set_probability_threshold(probability_threshold: float) → None

actableai.classification.roc_curve_cross_validation module

actableai.classification.roc_curve_cross_validation.cross_validation_curve(cross_val_auc_curves: Dict, x: str = 'False Positive Rate', y: str = 'True Positive Rate', negative_label: bool = True) → Dict

Computes the combined curves for ROC and Precision-Recall curves when using

cross-validation.

Parameters

roc_curves_dictionnary – A dictionnary containing the ROC curves for each classifier

Returns

A dictionnary containing the combined ROC curves for each classifier

actableai.classification.utils module

actableai.classification.utils.leaderboard_cross_val(cross_val_leaderboard: List[pandas.core.frame.DataFrame]) → pandas.core.frame.DataFrame

Creates a leaderboard from a list of cross validation results.

Parameters

cross_val_leaderboard – List of cross validation results.

Returns

Leaderboard.

Return type

pd.DataFrame

actableai.classification.utils.split_validation_by_datetime(df_train: pandas.core.frame.DataFrame, datetime_column: str, validation_ratio: float = 0.2) → Tuple[pandas.core.frame.DataFrame, pandas.core.frame.DataFrame]

Module contents