def featurize_dataframe(self, df, col_id, ignore_errors=False,
return_errors=False, inplace=True,
multiindex=False):
"""
Featurize dataframe is overloaded in order to allow
compatibility with Featurizers that overload featurize_dataframe
"""
if multiindex:
if not isinstance(df.columns, pd.MultiIndex):
col_id = ("Input Data", col_id)
df = homogenize_multiindex(df, "Input Data")
# Detect if any featurizers override featurize_dataframe
override = ["featurize_dataframe" in f.__class__.__dict__.keys()
for f in self.featurizers]
if any(override):
warnings.warn(
"One or more featurizers overrides featurize_dataframe, "
"featurization will be sequential and may diminish performance")
for f in self.featurizers:
df = f.featurize_dataframe(df, col_id, ignore_errors,
return_errors, inplace, multiindex)
if multiindex:
feature_labels = [(f.__class__.__name__, flabel) for flabel in f.feature_labels()]
else:
feature_labels = f.feature_labels()
df[feature_labels] = df[feature_labels].applymap(np.squeeze)
return df
def featurize_dataframe(self, df, col_id, *args, **kwargs):
"""
Accepts the same arguments as BaseFeaturizer.featurize_dataframe.
"""
multiindex = kwargs.get('multiindex', False)
if multiindex:
if not isinstance(df.columns, pd.MultiIndex):
col_id = ("Input Data", col_id)
df = homogenize_multiindex(df, "Input Data")
for f in self.featurizers:
df = f.featurize_dataframe(df, col_id, *args, **kwargs)
if multiindex:
feature_labels = [(f.__class__.__name__, flabel)
for flabel in f.feature_labels()]
else:
feature_labels = f.feature_labels()
df[feature_labels] = df[feature_labels].applymap(np.squeeze)
return df
def featurize_dataframe(self,
df,
col_id,
ignore_errors=False,
return_errors=False,
inplace=True,
multiindex=False,
pbar=True):
"""
Compute features for all entries contained in input dataframe.
Args:
df (Pandas dataframe): Dataframe containing input data.
col_id (str or list of str): column label containing objects to
featurize. Can be multiple labels if the featurize function
requires multiple inputs.
ignore_errors (bool): Returns NaN for dataframe rows where
exceptions are thrown if True. If False, exceptions
are thrown as normal.
return_errors (bool). Returns the errors encountered for each
row in a separate `XFeaturizer errors` column if True. Requires
ignore_errors to be True.
inplace (bool): Whether to add new columns to input dataframe (df)
multiindex (bool): If True, use a Featurizer - Feature 2-level
index using the MultiIndex capabilities of pandas. If done
inplace, multiindex featurization will overwrite the original
dataframe's column index.
pbar (bool): Shows a progress bar if True.
Returns:
updated dataframe.
"""
# If only one column and user provided a string, put it inside a list
if isinstance(col_id, string_types):
col_id = [col_id]
# Multiindexing doesn't play nice with other options!
if multiindex:
if inplace:
warnings.warn("Multiindexing enabled with inplace=True! The "
"original dataframe index has changed.")
elif isinstance(df.columns, pd.MultiIndex):
# If input df is multi, but multi not enabled...
raise ValueError(
"Please enable multiindexing to featurize an input"
" dataframe containing a column multiindex.")
# Generate the labels for the columns
labels = self._generate_column_labels(multiindex, return_errors)
# Check names to avoid overwriting the current columns
# ConversionFeaturizer have attribute called _overwrite_data which
# determines whether an Error is thrown
if not getattr(self, '_overwrite_data', False):
for col in df.columns.values:
if col in labels:
raise ValueError(
'"{}" exists in input dataframe'.format(col))
# Compute the features
features = self.featurize_many(df[col_id].values,
ignore_errors=ignore_errors,
return_errors=return_errors,
pbar=pbar)
# Make sure the dataframe can handle multiindices
if multiindex:
df = homogenize_multiindex(df, "Input Data")
# Create dataframe with the new features
res = pd.DataFrame(features, index=df.index, columns=labels)
if inplace:
# Update the existing dataframe
for k in labels:
df[k] = res[k]
return df
else:
# Create new dataframe and ensure columns are ordered properly
new = pd.concat([df, res], axis=1)
return new[df.columns.tolist() + res.columns.tolist()]
def featurize_dataframe(self,
df,
col_id,
ignore_errors=False,
return_errors=False,
inplace=True,
multiindex=False):
"""
Compute features for all entries contained in input dataframe.
Args:
df (Pandas dataframe): Dataframe containing input data.
col_id (str or list of str): column label containing objects to
featurize. Can be multiple labels if the featurize function
requires multiple inputs.
ignore_errors (bool): Returns NaN for dataframe rows where
exceptions are thrown if True. If False, exceptions
are thrown as normal.
return_errors (bool). Returns the errors encountered for each
row in a separate `XFeaturizer errors` column if True. Requires
ignore_errors to be True.
inplace (bool): Whether to add new columns to input dataframe (df)
Returns:
updated dataframe.
"""
# If only one column and user provided a string, put it inside a list
if isinstance(col_id, string_types):
col_id = [col_id]
# Generate the feature labels
labels = self.feature_labels()
# Check names to avoid overwriting the current columns
for col in df.columns.values:
if col in labels:
raise ValueError('"{}" exists in input dataframe'.format(col))
# Compute the features
features = self.featurize_many(df[col_id].values,
ignore_errors=ignore_errors,
return_errors=return_errors)
if return_errors:
labels.append(self.__class__.__name__ + " Exceptions")
if multiindex:
indices = ([self.__class__.__name__], labels)
labels = pd.MultiIndex.from_product(indices)
df = homogenize_multiindex(df, "Input Data")
elif isinstance(df.columns, pd.MultiIndex):
# If input df is multi, but multi not enabled...
raise ValueError(
"Please enable multiindexing to featurize an input"
" dataframe containing a column multiindex.")
# Create dataframe with the new features
res = pd.DataFrame(features, index=df.index, columns=labels)
if inplace:
# Update the existing dataframe
for k in labels:
df[k] = res[k]
return df
else:
# Create new dataframe and ensure columns are ordered properly
new = pd.concat([df, res], axis=1)
return new[df.columns.tolist() + res.columns.tolist()]
def featurize_dataframe(self, df, col_id, ignore_errors=False,
return_errors=False, inplace=True,
multiindex=False, pbar=True):
"""
Compute features for all entries contained in input dataframe.
Args:
df (Pandas dataframe): Dataframe containing input data.
col_id (str or list of str): column label containing objects to
featurize. Can be multiple labels if the featurize function
requires multiple inputs.
ignore_errors (bool): Returns NaN for dataframe rows where
exceptions are thrown if True. If False, exceptions
are thrown as normal.
return_errors (bool). Returns the errors encountered for each
row in a separate `XFeaturizer errors` column if True. Requires
ignore_errors to be True.
inplace (bool): Whether to add new columns to input dataframe (df)
multiindex (bool): If True, use a Featurizer - Feature 2-level
index using the MultiIndex capabilities of pandas. If done
inplace, multiindex featurization will overwrite the original
dataframe's column index.
pbar (bool): Shows a progress bar if True.
Returns:
updated dataframe.
"""
# If only one column and user provided a string, put it inside a list
if isinstance(col_id, string_types):
col_id = [col_id]
# Multiindexing doesn't play nice with other options!
if multiindex:
if inplace:
warnings.warn("Multiindexing enabled with inplace=True! The "
"original dataframe index has changed.")
elif isinstance(df.columns, pd.MultiIndex):
# If input df is multi, but multi not enabled...
raise ValueError("Please enable multiindexing to featurize an input"
" dataframe containing a column multiindex.")
# Generate the labels for the columns
labels = self._generate_column_labels(multiindex, return_errors)
# Check names to avoid overwriting the current columns
# ConversionFeaturizer have attribute called _overwrite_data which
# determines whether an Error is thrown
if not getattr(self, '_overwrite_data', False):
for col in df.columns.values:
if col in labels:
raise ValueError(
'"{}" exists in input dataframe'.format(col))
# Compute the features
features = self.featurize_many(df[col_id].values,
ignore_errors=ignore_errors,
return_errors=return_errors,
pbar=pbar)
# Make sure the dataframe can handle multiindices
if multiindex:
df = homogenize_multiindex(df, "Input Data")
# Create dataframe with the new features
res = pd.DataFrame(features, index=df.index, columns=labels)
if inplace:
# Update the existing dataframe
for k in labels:
df[k] = res[k]
return df
else:
# Create new dataframe and ensure columns are ordered properly
new = pd.concat([df, res], axis=1)
return new[df.columns.tolist() + res.columns.tolist()]