acfx.evaluation.ccfs

acfx.evaluation.ccfs.generate_cfs(query_instance: ndarray, desired_class: int, adjacency_matrix: ndarray, causal_order: Sequence[int], proximity_weight: float, sparsity_weight: float, plausibility_weight: float, diversity_weight: float, bounds: Dict[str, Tuple[float, float]], model: ClassifierMixin, features_order: List[str] | None = None, masked_features: List[str] | None = None, categorical_indicator: List[bool] | None = None, X: DataFrame | None = None, num_cfs: int = 1, init_points: int = 10, n_iter: int = 1000, sampling_from_model: bool = False, optimizer_type: OptimizerType = OptimizerType.EBM, optimizer: ModelBasedCounterOptimizer = None) → ndarray

Generate multiple counterfactuals that minimize the loss function using Bayesian Optimization.

Parameters:

query_instance:

The instance to generate counterfactuals for.

desired_class:

The target class for the counterfactuals.

adjacency_matrix:

The adjacency matrix representing the causal structure.

causal_order:

The order of variables in the causal graph.

proximity_weight:

Weight for proximity loss component

sparsity_weight:

Weight for sparsity loss component

plausibility_weight:

Weight for plausibility loss component

diversity_weight:

Weight for diversity loss component

bounds:

The bounds for each feature to search over (dict with feature names as keys and tuple (min, max) as values).

model:

The predictive model used to predict class labels.

features_order:

order of features in query instance

masked_features:

List of interchangeable features

categorical_indicator:

True at the index where the variable should be treated as categorical

num_cfs:

The number of counterfactual instances to generate.

init_points:

Number of initial points for Bayesian Optimization.

n_iter:

Number of iterations for Bayesian Optimization.

X:

training dataset to sample from.

sampling_from_model:

true if you want to generate samples from model after sampling from data

optimizer_type:

type of optimizer used on model to generate counterfactuals. If you use OptimizerType.Custom, you need to provide optimizer instance

optimizer:

instance of optimizer (use for optimizer_type = OptimizerType.Custom)

Returns

np.ndarray:

The generated counterfactuals that minimize the loss function.

acfx.evaluation.ccfs.generate_cfs_bayesian(query_instance: ndarray, desired_class: int, bayesian_model: DiscreteBayesianNetwork, categorical_bins_num: int, proximity_weight: float, sparsity_weight: float, plausibility_weight: float, diversity_weight: float, bounds: Dict[str, Tuple[float, float]], model: ClassifierMixin, features_order: List[str] | None = None, masked_features: List[str] | None = None, categorical_indicator: List[bool] | None = None, X: DataFrame | None = None, num_cfs: int = 1, init_points: int = 10, n_iter: int = 1000, sampling_from_model: bool = False, optimizer_type: OptimizerType = OptimizerType.EBM, optimizer: ModelBasedCounterOptimizer = None) → ndarray

Generate multiple counterfactuals that minimize the loss function using Bayesian Optimization and bayesian network to calculate causal loss.

Parameters:

query_instance:

The instance to generate counterfactuals for.

desired_class:

The target class for the counterfactuals.

bayesian_model:

Fitted bayesian network

categorical_bins_num:

Number of bins that were used to discretize continuous features.

proximity_weight:

Weight for proximity loss component

sparsity_weight:

Weight for sparsity loss component

plausibility_weight:

Weight for plausibility loss component

diversity_weight:

Weight for diversity loss component

bounds:

The bounds for each feature to search over (dict with feature names as keys and tuple (min, max) as values).

model:

The predictive model used to predict class labels.

features_order:

order of features in query instance

masked_features:

List of interchangeable features

categorical_indicator:

True at the index where the variable should be treated as categorical

num_cfs:

The number of counterfactual instances to generate.

init_points:

Number of initial points for Bayesian Optimization.

n_iter:

Number of iterations for Bayesian Optimization.

X:

training dataset to sample from.

sampling_from_model:

true if you want to generate samples from model after sampling from data

optimizer_type:

type of optimizer used on model to generate counterfactuals. If you use OptimizerType.Custom, you need to provide optimizer instance

optimizer:

instance of optimizer (use for optimizer_type = OptimizerType.Custom)

Returns

np.ndarray:

The generated counterfactuals that minimize the loss function.