def integer_to_count(cls) -> InferenceRelation:
return InferenceRelation(
relationship=is_unsigned_int,
transformer=lambda s: s.astype(np.uint64),
type=cls,
related_type=Integer,
)
def integer_to_count(cls) -> InferenceRelation:
return InferenceRelation(
relationship=is_unsigned_int,
transformer=to_unsigned_int,
type=cls,
related_type=Integer,
)
def categorical_to_ordinal(cls) -> InferenceRelation:
from visions.types import Categorical
return InferenceRelation(cls,
Categorical,
relationship=is_ordinal_cat,
transformer=to_ordinal)
def string_to_ordinal(cls) -> InferenceRelation:
from visions.types import String
return InferenceRelation(
type=cls,
related_type=String,
relationship=is_ordinal_str,
transformer=to_ordinal,
)
def string_to_categorical_distinct_count(cls) -> InferenceRelation:
"""Convert string to categorical when it has fewer than 50% unique values.
Returns:
relation
"""
# TODO: only when not any other string relation (either exclude others or have ordering and evaluate last)
return InferenceRelation(
relationship=lambda s, state: s.nunique() / len(s) < 0.5,
transformer=lambda s: s.astype("category"),
related_type=String,
type=cls,
)