from sklearn.impute import SimpleImputer as SimpleImputerOperation
from DashAI.back.api.utils import cast_string_to_type
from DashAI.back.converters.sklearn_wrapper import SklearnWrapper
from DashAI.back.core.schema_fields import (
bool_field,
enum_field,
float_field,
int_field,
none_type,
schema_field,
string_field,
union_type,
)
from DashAI.back.core.schema_fields.base_schema import BaseSchema
class SimpleImputerSchema(BaseSchema):
missing_values: schema_field(
none_type(
union_type(int_field(), union_type(float_field(), string_field()))
), # int, float, str, np.nan, None or pandas.NA
None, # np.nan,
"The placeholder for the missing values.",
) # type: ignore
strategy: schema_field(
enum_field(["mean", "median", "most_frequent", "constant"]),
"mean",
"The imputation strategy.",
) # type: ignore
fill_value: schema_field(
none_type(union_type(int_field(), union_type(float_field(), string_field()))),
None,
"The value to replace missing values with.",
) # type: ignore
use_copy: schema_field(
bool_field(),
True,
"If True, a copy of X will be created.",
alias="copy",
) # type: ignore
add_indicator: schema_field(
bool_field(),
False,
"If True, a MissingIndicator transform will stack onto output.",
) # type: ignore
keep_empty_features: schema_field(
bool_field(),
False,
"If True, empty features will be kept.",
) # type: ignore
[docs]
class SimpleImputer(SklearnWrapper, SimpleImputerOperation):
"""SciKit-Learn's SimpleImputer wrapper for DashAI."""
SCHEMA = SimpleImputerSchema
DESCRIPTION = (
"Univariate imputer for completing missing values with simple strategies."
)
[docs]
def __init__(self, **kwargs):
self.missing_values = kwargs.pop("missing_values", None)
self.missing_values = cast_string_to_type(self.missing_values)
kwargs["missing_values"] = self.missing_values
super().__init__(**kwargs)