def test_multi_label_encoder(self):
df = pd.DataFrame({"A": [1, 2, 3, 4], "B": ['a', 'a', 'a', 'b']})
df_expect = pd.DataFrame({"A": [0, 1, 2, 3], "B": [0, 0, 0, 1]})
ec = skex.MultiLabelEncoder(dtype=np.int64)
df_t = ec.fit_transform(df.copy())
assert np.where(df_expect.values == df_t.values, 0, 1).sum() == 0
assert all(df_t.dtypes == pd.Series(dict(A=np.int64, B=np.int64)))
ec = skex.MultiLabelEncoder(dtype=np.int32)
df_t = ec.fit_transform(df.copy())
assert np.where(df_expect.values == df_t.values, 0, 1).sum() == 0
assert all(df_t.dtypes == pd.Series(dict(A=np.int32, B=np.int32)))
def _categorical_encoding(self, X):
start = time.time()
logger.info('Categorical encoding...')
vars = self.get_categorical_columns()
mle = skex.MultiLabelEncoder(vars)
X = mle.fit_transform(X)
self.X_transformers['label_encoder'] = mle
print(f'Categorical encoding taken {time.time() - start}s')
return X
def test_func_transformer(self):
dfm = DataFrameMapper(
[(column_object_category_bool, [
SimpleImputer(strategy='constant'),
skex.MultiLabelEncoder(),
]
),
],
input_df=True,
df_out=True,
df_out_dtype_transforms=[
(column_object, 'category')
]
)
X, y = get_df()
x_new = dfm.fit_transform(X, y)
assert x_new.dtypes.to_list() == [pd.CategoricalDtype(categories=[0, 1, 2], ordered=False),
pd.CategoricalDtype(categories=[0, 1], ordered=False),
pd.CategoricalDtype(categories=[0, 1, 2], ordered=False)]
def test_cache():
clear()
df = dsutils.load_bank()
t = skex.MultiLabelEncoder()
X = t.fit_transform(df.copy())
t1 = CachedMultiLabelEncoder()
X1 = t1.fit_transform(df.copy())
t2 = CachedMultiLabelEncoder()
X2 = t2.fit_transform(df.copy())
hasher = get_tool_box(df).data_hasher()
assert hasher(X) == hasher(X1) == hasher(X2)
t3 = CachedMultiLabelEncoder()
X3 = t3.fit_transform_as_tuple_result(df.copy())
t4 = CachedMultiLabelEncoder()
X4 = t4.fit_transform_as_tuple_result(df.copy())
assert isinstance(X3, (tuple, list))
assert isinstance(X4, (tuple, list))
assert hasher(X3[1]) == hasher(X4[1])
def as_local(self):
target = sk_ex.MultiLabelEncoder()
target.columns = self.columns
target.dtype = self.dtype
target.encoders = {k: e.as_local() for k, e in self.encoders.items()}
return target
def load_data():
df = dsutils.load_bank()
return skex.MultiLabelEncoder().fit_transform(df)
