Evaluation metrics

class evaluation.evaluation_object.EvaluationObject(**kwargs)

Bases: ABC

Base class for a metric comparing factual and counterfactual datasets.

Each evaluation consumes the finalized factual dataset and the counterfactual dataset returned by predict(), and returns a one-row dataframe of named metrics. Experiment concatenates these one-row frames column-wise into the final metrics table.

abstractmethod evaluate(factuals, counterfactuals)

Compute the metric.

Parameters:

• factuals (DatasetObject) – The finalized factual dataset.

• counterfactuals (DatasetObject) – The counterfactual dataset. If it carries an evaluation_filter, apply it before computing the metric.

Returns:

A single-row dataframe with stable, descriptive column names.

Return type:

pandas.DataFrame