def _transform(self, X: DataFrameContainer, y: Optional[NdArrayContainer]):
y_data = y.data
X_data, y_data = self._transform_proc(X.data, y_data)
X = X.copy()
y = y.copy()
X_data = pd.DataFrame(X_data, columns=X.columns)
X.data = X_data
y.data = y_data
return X, y
def process_X(self, X: DataFrameContainer, X_origin):
if X is None:
return None
assert X.shape[1] == len(self.columns)
if isinstance(X_origin, np.ndarray):
X.columns = self.columns
elif isinstance(X_origin, pd.DataFrame):
assert set(X.columns) == set(self.columns)
if not np.all(X.columns == self.columns):
self.logger.warning(
f"{X.dataset_source}'s columns do not match the TrainSet's columns by position!"
)
X.data = X.data[self.columns]
elif isinstance(X_origin, DataFrameContainer):
pass
else:
raise NotImplementedError
X.set_feature_groups(self.feature_groups)
return X