hoeffding_tree

HoeffdingTree(inputs, seed, model_handler, specs_handler)

Train a Hoeffding Tree on synthetic samples.

Samples are generated from another model.

Attributes:

datamodel: DataModel Object used to generate synthetic training inputs based on the original dataset.

list

Original training inputs transformed to River-compatible format with predictions.

list

List of indices for nominal features.

Methods:

train_tree() Trains a Hoeffding Tree and returns the trained model.

Initializes the HoeffdingTree trainer.

Parameters:

Name	Type	Description	Default
inputs	DataFrame	Original training dataset including target column.	required
seed	int	Random seed for reproducible synthetic sample generation.	required
model_handler	ModelHandler	Object used to generate predictions from the Flowcean model.	required
specs_handler	SystemSpecsHandler	Object containing feature specifications and metadata.	required

Source code in src/flowcean/testing/generator/ddtig/domain/model_analyser/mut/hoeffding_tree.py

def __init__(
    self,
    inputs: pl.DataFrame,
    seed: int,
    model_handler: ModelHandler,
    specs_handler: SystemSpecsHandler,
) -> None:
    """Initializes the HoeffdingTree trainer.

    Args:
        inputs: Original training dataset including target column.
        seed: Random seed for reproducible synthetic sample generation.
        model_handler: Object used to generate predictions from
            the Flowcean model.
        specs_handler: Object containing feature specifications
            and metadata.
    """
    # Remove target column to isolate input features
    inputs = inputs.drop(inputs.columns[-1])
    self.datamodel = DataModel(inputs, seed, model_handler, specs_handler)

    # Generate River-compatible samples using original data
    self.samples = self.datamodel.generate_dataset(original_data=True)
    self.nominal_attributes = specs_handler.get_nominal_features()

train_tree(performance_threshold, sample_limit, n_predictions, *, classification, **kwargs)

Train a Hoeffding Tree using synthetic samples.

Continue until performance criteria are met.

Parameters:

Name	Type	Description	Default
performance_threshold	float	Minimum performance required to finalize the model.	required
sample_limit	int	Maximum number of samples to use during training.	required
n_predictions	int	Number of consecutive correct predictions required to stop training.	required
classification	bool	Indicates whether the task is classification or regression.	required
**kwargs	Any	Additional hyperparameters for the Hoeffding Tree model.	{}

Returns:

Type	Description
HoeffdingTreeRegressor | HoeffdingTreeClassifier	Trained Hoeffding Tree model.

Source code in src/flowcean/testing/generator/ddtig/domain/model_analyser/mut/hoeffding_tree.py

def train_tree(
    self,
    performance_threshold: float,
    sample_limit: int,
    n_predictions: int,
    *,
    classification: bool,
    **kwargs: Any,
) -> HoeffdingTreeRegressor | HoeffdingTreeClassifier:
    """Train a Hoeffding Tree using synthetic samples.

    Continue until performance criteria are met.

    Args:
        performance_threshold: Minimum performance required to
            finalize the model.
        sample_limit: Maximum number of samples to use during training.
        n_predictions: Number of consecutive correct predictions
            required to stop training.
        classification: Indicates whether the task is
            classification or regression.
        **kwargs: Additional hyperparameters for the Hoeffding Tree model.

    Returns:
        Trained Hoeffding Tree model.
    """
    metric, model = self._create_model_and_metric(
        classification=classification,
        **kwargs,
    )

    # Pre-train
    for x, y in self.samples:
        y_true = self._normalize_target(y, classification=classification)
        model.learn_one(x, y_true)

    self._run_training_loop(
        model=model,
        metric=metric,
        performance_threshold=performance_threshold,
        n_predictions=n_predictions,
        sample_limit=sample_limit,
        classification=classification,
    )

    logger.info("Hoeffding Tree training completed successfully.")
    return model