def test_robust_imputer_categorical_custom_function():
robust_imputer = RobustImputer(
dtype=np.dtype("O"), strategy="constant", fill_values="not hot dog", mask_function=lambda x: x == "hot dog"
)
robust_imputer.fit(X_impute_categorical)
X_observed = robust_imputer.transform(X_impute_categorical)
assert_array_equal(X_observed, X_imputed_categorical)
def test_robust_imputer_transform_dim_error():
with pytest.raises(ValueError, match=transform_error_msg):
robust_imputer = RobustImputer()
robust_imputer.fit(X_impute)
robust_imputer.transform(np.zeros((3, 4)))
def test_robust_imputer(X, X_expected, strategy, fill_values):
robust_imputer = RobustImputer(strategy=strategy, fill_values=fill_values)
robust_imputer.fit(X)
X_observed = robust_imputer.transform(X)
assert_array_equal(X_observed, X_expected)
class NALabelEncoder(BaseEstimator, TransformerMixin):
"""Encoder for transforming labels to NA values.
Uses `RobustImputer` on 1D inputs of labels
- Uses `is_finite_numeric` mask for encoding by default
- Only uses the `RobustImputer` strategy `constant` and fills using `np.nan`
- Default behavior encodes non-float and non-finite values as nan values in
the target column of a given regression dataset
Parameters
----------
mask_function : callable -> np.array, dtype('bool') (default=None)
A vectorized python function, accepts np.array, returns np.array
with dtype('bool')
For each value, if mask_function(val) == False, that value will
be imputed. mask_function is used to create a boolean mask that determines
which values in the input to impute.
Use np.vectorize to vectorize singular python functions.
"""
def __init__(self, mask_function=None):
self.mask_function = mask_function
def fit(self, y):
"""Fit the encoder on y.
Parameters
----------
y : {array-like}, shape (n_samples,)
Input column, where `n_samples` is the number of samples.
Returns
-------
self : NALabelEncoder
"""
self.model_ = RobustImputer(strategy="constant",
fill_values=np.nan,
mask_function=self.mask_function)
y = y.reshape(-1, 1)
self.model_.fit(X=y)
return self
def transform(self, y):
"""Encode all non-float and non-finite values in y as NA values.
Parameters
----------
y : {array-like}, shape (n_samples)
The input column to encode.
Returns
-------
yt : {ndarray}, shape (n_samples,)
The encoded input column.
"""
check_is_fitted(self, "model_")
y = y.reshape(-1, 1)
return self.model_.transform(y).flatten()
def inverse_transform(self, y):
"""Returns input column"""
return y
def _more_tags(self):
return {"X_types": ["1dlabels"]}