autoprognosis.plugins.uncertainty.plugin_cohort_explainer module

class CohortExplainerPlugin(model: Any, task_type: str = 'classification', random_seed: int = 0, effect_size: float = 0.5)

Bases: UncertaintyPlugin

Uncertainty plugin based on Conformal Prediction and cohort analysis.

Parameters:

• model – model. The model to explain.

• task_type – str risk_estimation, regression, classification.

• random_seed – int Random seed

• effect_size – float Effect size for the risk effect size explainer.

fit(*args: Any, **kwargs: Any) → UncertaintyPlugin

static name() → str

predict(*args: Any, **kwargs: Any) → DataFrame

predict_proba(*args: Any, **kwargs: Any) → DataFrame

static type() → str

class CohortMgmt(cohort_rules: Dict[str, CohortRule])

Bases: object

diagnostics(X: DataFrame, prediction: DataFrame, cohort_limit: int = 2) → dict

diagnostics_df(X: DataFrame, prediction: DataFrame) → dict

diagnostics_headers() → list

get(name: str) → CohortRule | None

global_cohort = '(*)'

match(X: DataFrame) → CohortRule | None

match_all(X: DataFrame) → List[CohortRule]

class CohortRule(name: str, target_features: list = [], ops: list = [], thresholds: list = [])

Bases: object

calibrate(ground_truth: ndarray, predictions: ndarray) → bool

get_confidence(prediction: ndarray) → ndarray

get_difficulty() → float

match(X: DataFrame) → DataFrame

merge(other: CohortRule) → CohortRule

plugin

alias of CohortExplainerPlugin