def gen_groupby_cat_encode_features(self, data, cat_columns, num_column,
cat_encoder_name='JamesSteinEncoder'):
"""
Description of group_encoder
Args:
data (pd.DataFrame): dataset
cat_columns (list): cat columns names
num_column (str): num column name
Returns:
pd.DataFrame
"""
if isinstance(cat_encoder_name, str):
if cat_encoder_name in self.cat_encoder_names_list:
encoder = JamesSteinEncoder(cols=self.cat_features, model='beta', return_df = True, drop_invariant=True)
encoder.fit(X=data[cat_columns], y=data[num_column].values)
else:
print(f"{cat_encoder_name} is not supported!")
return ('', '')
else:
encoder = copy.deepcopy(cat_encoder_name)
data_encodet = encoder.transform(X=data[cat_columns], y=data[num_column].values)
data_encodet = data_encodet.add_prefix('GroupEncoded_' + num_column + '_')
return (data_encodet, encoder)
def JamesStein_Encoding(self,
model: str = 'independent',
sigma: float = 0.05,
randomized: bool = False):
"""
James-Stein编码，也是一种基于目标编码的编码方法，也尝试通过参数B来平衡先验概率与观测到的条件概率。
但与目标编码与M估计量编码不同的是，James-Stein编码器通过方差比而不是样本大小来平衡两个概率。
:param model:
:param sigma:
:param randomized:
:return:
"""
self.encoder = JamesSteinEncoder(cols=self.cols,
model=model,
sigma=sigma,
randomized=randomized)
TargetClassifierEncoderCV,
TargetRegressorEncoder,
TargetRegressorEncoderCV,
)
from category_encoders import JamesSteinEncoder
from sklearn.preprocessing import LabelEncoder
DATA_PATH = Path(".") / "data"
@pytest.mark.parametrize(
"encoder",
[
TargetClassifierEncoder(),
TargetClassifierEncoderCV(),
JamesSteinEncoder(),
],
)
def test_adult(encoder):
"""Smoke test for adult dataset."""
adult_path = DATA_PATH / "adult.csv"
adult_df = pd.read_csv(adult_path)
X = adult_df.drop("class", axis=1)
y = LabelEncoder().fit_transform(adult_df["class"])
prep = ColumnTransformer([
(
"cat",
encoder,
}
ENCODERS = {
"drop":
"drop",
"SKOrdinalEncoder":
make_pipeline(
SimpleImputer(strategy="constant", fill_value="sk_missing"),
OrdinalEncoder(handle_unknown="use_encoded_value", unknown_value=-1),
),
"SKTargetEncoder":
TargetRegressorEncoder(),
"SKTargetEncoderCV":
TargetRegressorEncoderCV(),
"JamesSteinEncoder":
JamesSteinEncoder(),
"JamesSteinEncoderCV":
NestedEncoderCV(JamesSteinEncoder()),
}
def run_single_benchmark(data_str, encoder_str, cv, n_jobs, write_result,
force):
print(f"running benchmark for {data_str} and {encoder_str}")
data_info = DATA_INFOS[data_str]
encoder = ENCODERS[encoder_str]
results_path = get_results_path(RESULTS_DIR, data_info, encoder_str)
if results_path.exists() and not force:
print(
f"benchmark for {data_str} and {encoder_str} exists pass --force to rerun"