class OrdinalEncoderImpl():
def __init__(self, categories='auto', dtype=None):
self._hyperparams = {'categories': categories, 'dtype': dtype}
def fit(self, X, y=None):
self._sklearn_model = SKLModel(**self._hyperparams)
if (y is not None):
self._sklearn_model.fit(X, y)
else:
self._sklearn_model.fit(X)
return self
def transform(self, X):
return self._sklearn_model.transform(X)
class NanOrdinalEncoder(_BaseEncoder):
def __init__(self):
self.lbe = OrdinalEncoder()
def fit(self, X):
self.lbe.fit(X)
return self
def transform(self, X):
result = self.lbe.transform(X)
for col in range(result.shape[1]):
nan_idx = list(self.lbe.categories_[col]).index(nan_replace)
column = result[:, col]
column[column == nan_idx] = np.nan
return result
def fit(self, X, y=None):
self._sklearn_model = SKLModel(**self._hyperparams)
if (y is not None):
self._sklearn_model.fit(X, y)
else:
self._sklearn_model.fit(X)
return self
def calculate_metafeatures(dataset, dataset_id=None, data_dir='./', task_type=None):
if isinstance(dataset, str):
X, y, feature_types = load_data(dataset, data_dir, datanode_returned=False, preprocess=False, task_type=task_type)
dataset_id = dataset
elif isinstance(dataset, DataNode):
X, y, feature_types = dataset.data[0], dataset.data[1], dataset.feature_types
import pandas as pd
X = pd.DataFrame(data=X)
else:
raise ValueError('Invalid dataset input!')
categorical_idx = [i for i, feature_type in enumerate(feature_types) if feature_type == 'categorical']
nan_val = np.array(X.isnull()).astype('int')
nan_avg = np.average(nan_val, axis=0)
nan_idx = [idx for idx in range(len(nan_avg)) if nan_avg[idx] != 0]
nan_categorical_idx = list(set(nan_idx).intersection(categorical_idx))
for col in X.columns[nan_categorical_idx]:
X[col].fillna(nan_replace, inplace=True)
X = np.array(X)
y = np.array(y)
normal_categorical_idx = list(set(categorical_idx) - set(nan_categorical_idx))
lbe = ColumnTransformer([('lbe', OrdinalEncoder(), normal_categorical_idx),
('nan_lbe', NanOrdinalEncoder(), nan_categorical_idx)],
remainder="passthrough")
X = lbe.fit_transform(X).astype('float64')
categorical_ = [True] * len(categorical_idx)
categorical_false = [False] * (len(feature_types) - len(categorical_idx))
categorical_.extend(categorical_false)
mf = calculate_all_metafeatures(X=X, y=y,
categorical=categorical_,
dataset_name=dataset_id,
task_type=task_type)
return mf.load_values()
def __init__(self):
self.lbe = OrdinalEncoder()
def __init__(self, categories='auto', dtype=None):
self._hyperparams = {'categories': categories, 'dtype': dtype}
self._wrapped_model = SKLModel(**self._hyperparams)