autoprognosis.plugins.ensemble.classifiers module

class AggregatingEnsemble(*args: Any, **kwargs: Any)

Bases: BaseEnsemble

Basic ensemble strategies:

• average: average across all scores/prediction results, maybe with weights

• maximization: simple combination by taking the maximum scores

• majority vote

• median: take the median value across all scores/prediction results

Parameters:

• models – list. List of base models.

• method – str. average, maximization, majority vote, median

• explainer_plugins – list. List of explainers attached to the ensemble.

enable_explainer(explainer_plugins: list = [], explanations_nepoch: int = 10000) → None

explain(X: DataFrame, *args: Any) → DataFrame

fit(X: DataFrame, Y: DataFrame) → AggregatingEnsemble

is_fitted() → bool

classmethod load(buff: bytes) → AggregatingEnsemble

name() → str

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

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

save() → bytes

score(X: DataFrame, y: DataFrame, metric: str = 'aucroc') → float

class BaseEnsemble(*args: Any, **kwargs: Any)

Bases: BaseEstimator

Abstract ensemble interface

enable_explainer(explainer_plugins: list = [], explanations_nepoch: int = 10000) → None

abstract explain(X: DataFrame, *args: Any) → DataFrame

abstract fit(X: DataFrame, Y: DataFrame) → BaseEnsemble

abstract is_fitted() → bool

abstract classmethod load(buff: bytes) → BaseEnsemble

abstract name() → str

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

abstract predict_proba(X: DataFrame, *args: Any) → DataFrame

abstract save() → bytes

score(X: DataFrame, y: DataFrame, metric: str = 'aucroc') → float

class StackingEnsemble(*args: Any, **kwargs: Any)

Bases: BaseEnsemble

Stacking ensemble(meta ensembling): Use a meta-learner on top of the base models

Parameters:

• models – list. List of base models.

• meta_model – Pipeline. The meta learner.

• explainer_plugins – list. List of explainers attached to the ensemble.

enable_explainer(explainer_plugins: list = [], explanations_nepoch: int = 10000) → None

explain(X: DataFrame, *args: Any) → DataFrame

fit(X: DataFrame, Y: DataFrame) → StackingEnsemble

is_fitted() → bool

classmethod load(buff: bytes) → StackingEnsemble

name() → str

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

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

save() → bytes

score(X: DataFrame, y: DataFrame, metric: str = 'aucroc') → float

class WeightedEnsemble(*args: Any, **kwargs: Any)

Bases: BaseEnsemble

Weighted ensemble

Parameters:

• models – list. List of base models.

• weights – list. The weights for each base model.

• explainer_plugins – list. List of explainers attached to the ensemble.

enable_explainer(explainer_plugins: list = [], explanations_nepoch: int = 10000) → None

explain(X: DataFrame, *args: Any) → DataFrame

fit(X: DataFrame, Y: DataFrame) → WeightedEnsemble

is_fitted() → bool

classmethod load(buff: bytes) → WeightedEnsemble

name() → str

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

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

save() → bytes

score(X: DataFrame, y: DataFrame, metric: str = 'aucroc') → float

class WeightedEnsembleCV(*args: Any, **kwargs: Any)

Bases: BaseEnsemble

Cross-validated Weighted ensemble, with uncertainity prediction support

Parameters:

• models – list. List of base models.

• weights – list. The weights for each base model.

• explainer_plugins – list. List of explainers attached to the ensemble.

enable_explainer(explainer_plugins: list = [], explanations_nepoch: int = 10000) → None

explain(X: DataFrame, *args: Any) → DataFrame

fit(X: DataFrame, Y: DataFrame) → WeightedEnsembleCV

is_fitted() → bool

classmethod load(buff: bytes) → WeightedEnsembleCV

name() → str

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

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

predict_proba_with_uncertainity(X: DataFrame, *args: Any) → Tuple[DataFrame, DataFrame]

save() → bytes

score(X: DataFrame, y: DataFrame, metric: str = 'aucroc') → float