import pyarrow as pa
from sklearn.feature_selection import (
GenericUnivariateSelect as GenericUnivariateSelectOperation,
)
from DashAI.back.converters.category.feature_selection import FeatureSelectionConverter
from DashAI.back.converters.sklearn_wrapper import SklearnWrapper
from DashAI.back.core.schema_fields import (
enum_field,
float_field,
int_field,
none_type,
schema_field,
union_type,
)
from DashAI.back.core.schema_fields.base_schema import BaseSchema
from DashAI.back.types.dashai_data_type import DashAIDataType
from DashAI.back.types.value_types import Float
class GenericUnivariateSelectSchema(BaseSchema):
mode: schema_field(
enum_field(["percentile", "k_best", "fpr", "fdr", "fwe"]),
"percentile",
"Select features according to a percentile of the highest scores.",
) # type: ignore
param: schema_field(
none_type(
union_type(enum_field(["all"]), union_type(float_field(), int_field()))
),
1e-5,
"Parameter of the mode.",
) # type: ignore
[docs]
class GenericUnivariateSelect(
FeatureSelectionConverter, SklearnWrapper, GenericUnivariateSelectOperation
):
"""SciKit-Learn's GenericUnivariateSelect wrapper for DashAI."""
SCHEMA = GenericUnivariateSelectSchema
DESCRIPTION = "Univariate feature selector with configurable strategy."
SUPERVISED = True
DISPLAY_NAME = "Generic Univariate Select"
IMAGE_PREVIEW = "generic_univariate_select.png"
metadata = {}
def get_output_type(self, column_name: str = None) -> DashAIDataType:
"""Returns Float64 as the output type for selected features."""
return Float(arrow_type=pa.float64())
CATEGORY = "Feature Selection"