data_model

DataModel(data, seed, model_handler, specs_handler)

Generate synthetic samples from the training data distribution.

Attributes:

data: pl.DataFrame Original training data used in the Flowcean model.

list

Names of the columns in the training data.

int

Number of features in the dataset.

ModelHandler

ModelHandler object used to produce predictions.

list

List of indices for features of type int.

Methods:

generate_dataset() Generate random samples based on data distribution, or use original data.

Initializes the DataModel.

Parameters:

Name	Type	Description	Default
data	DataFrame	Original training data used in the Flowcean model.	required
seed	int	Random seed for reproducibility.	required
model_handler	ModelHandler	ModelHandler object used to produce predictions.	required
specs_handler	SystemSpecsHandler	SystemSpecsHandler object storing system specifications.	required

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

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

    Args:
        data: Original training data used in the Flowcean model.
        seed: Random seed for reproducibility.
        model_handler: ModelHandler object used to produce predictions.
        specs_handler: SystemSpecsHandler object storing
            system specifications.
    """
    self.data = data
    self.seed = seed
    self.col_names = data.columns
    self.model_handler = model_handler

    self.n_features = specs_handler.get_n_features()
    self.int_features = specs_handler.get_int_features()

generate_dataset(*, original_data=False, n_samples=0)

Generates a dataset of inputs and corresponding model predictions.

If original_data is True, uses the original training data. Otherwise, generates synthetic samples using KDE.

Parameters:

Name	Type	Description	Default
original_data	bool	Whether to use original training data or generate synthetic samples.	False
n_samples	int	Number of synthetic samples to generate.	0

Returns:

Name	Type	Description
	list	List of tuples containing input dictionaries and model outputs.
Example	n_samples = 1
	list	[({'Length': 0.5093, 'Diameter': 0.3886,
	list	'Height': 0.1106, 'M': 0}, 8.6006)]

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

def generate_dataset(
    self,
    *,
    original_data: bool = False,
    n_samples: int = 0,
) -> list:
    """Generates a dataset of inputs and corresponding model predictions.

    If original_data is True, uses the original training data.
    Otherwise, generates synthetic samples using KDE.

    Args:
        original_data: Whether to use original training
            data or generate synthetic samples.
        n_samples: Number of synthetic samples to generate.

    Returns:
        List of tuples containing input dictionaries and model outputs.
        Example (n_samples = 1):
        [({'Length': 0.5093, 'Diameter': 0.3886,
        'Height': 0.1106, 'M': 0}, 8.6006)]
    """
    training_inputs = (
        self.data
        if original_data
        else self._generate_samples(n_samples, self.int_features)
    )
    training_outputs = self.model_handler.get_model_prediction(
        training_inputs,
    ).collect()
    samples_input_lst = training_inputs.to_dicts()
    samples_output_lst = pl.Series(
        training_outputs.select(training_outputs.columns[0]),
    ).to_list()
    return [
        (inputs, output)
        for inputs, output in zip(
            samples_input_lst,
            samples_output_lst,
            strict=False,
        )
    ]