def local_transform(self, transformation, k, force_unknown=None):
'''
return a new AnonymDataFrame with a transformed self.df
df is modified by application of transformation
The main difference with transformation is that here
tranformation are applied by each group only if needed.
- transformation: can be
- a list of tuple with:
- first element is the name the column
- second element is the transformation
- no dict here as order counts
- k: un entier est le k-anonymat recherché
Note: it does have effect here but transformation are applied
in the self.variables order or in the order of list when
transformation is a list
'''
if force_unknown is None:
force_unknown = self.unknown
self.transformation = transformation
assert isinstance(transformation, list)
assert all([len(x) == 2 for x in transformation])
assert all([x[0] in self.df.columns for x in transformation])
variables = [x[0] for x in transformation]
derniere_transfo = transformation[-1]
anonymized_df = self.df.copy()
if get_k(anonymized_df, variables, force_unknown) >= k:
self.anonymized_df = anonymized_df
return self.copy()
if len(transformation) == 1:
colname = transformation[0][0]
transfo = transformation[0][1]
anonymized_df[colname] = transfo(anonymized_df[colname])
self.anonymized_df = anonymized_df
return self.copy()
if get_k(anonymized_df, variables[:-1], force_unknown) < k:
anonymized_df = self.local_transform(transformation[:-1],
k).anonymized_df
# on a une table k-anonymisée lorsqu'elle est restreinte aux
# len(variables) - 1 premières variables
# on applique l'aggrégation locale d'une variable par groupe
grp = anonymized_df.groupby(variables[:-1])
fonction = derniere_transfo[1]
variable = derniere_transfo[0]
anonymized_df[variable] = grp[variable].apply(fonction)
#assert get_k(anonymized_df, variables, force_unknown) >= k
self.anonymized_df = anonymized_df
return self.copy()
def local_transform(self, transformation, k, force_unknown=None):
'''
return a new AnonymDataFrame with a transformed self.df
df is modified by application of transformation
The main difference with transformation is that here
tranformation are applied by each group only if needed.
- transformation: can be
- a list of tuple with:
- first element is the name the column
- second element is the transformation
- no dict here as order counts
- k: un entier est le k-anonymat recherché
Note: it does have effect here but transformation are applied
in the self.variables order or in the order of list when
transformation is a list
'''
if force_unknown is None:
force_unknown = self.unknown
self.transformation = transformation
assert isinstance(transformation, list)
assert all([len(x) == 2 for x in transformation])
assert all([x[0] in self.df.columns for x in transformation])
variables = [x[0] for x in transformation]
derniere_transfo = transformation[-1]
anonymized_df = self.df.copy()
if get_k(anonymized_df, variables, force_unknown) >= k:
self.anonymized_df = anonymized_df
return self.copy()
if len(transformation) == 1:
colname = transformation[0][0]
transfo = transformation[0][1]
anonymized_df[colname] = transfo(anonymized_df[colname])
self.anonymized_df = anonymized_df
return self.copy()
if get_k(anonymized_df, variables[:-1], force_unknown) < k:
anonymized_df = self.local_transform(transformation[:-1], k).anonymized_df
# on a une table k-anonymisée lorsqu'elle est restreinte aux
# len(variables) - 1 premières variables
# on applique l'aggrégation locale d'une variable par groupe
grp = anonymized_df.groupby(variables[:-1])
fonction = derniere_transfo[1]
variable = derniere_transfo[0]
anonymized_df[variable] = grp[variable].apply(fonction)
#assert get_k(anonymized_df, variables, force_unknown) >= k
self.anonymized_df = anonymized_df
return self.copy()
def test_get_k(self):
iris = pd.read_csv("data/iris.csv")
k = get_k(iris, ['Name'])
self.assertEqual(k, 50)
def get_k(self):
return get_k(self.df, self.identifiant)
def get_final_k(self):
return get_k(self.anonymized_df, self.identifiant, self.unknown)
def get_k(self):
return get_k(self.df, self.identifiant, self.unknown)
def test_get_k(self):
iris = pd.read_csv("data/iris.csv")
k = get_k(iris, ['Name'])
self.assertEqual(k, 50)
def get_final_k(self):
return get_k(self.anonymized_df, self.identifiant, self.unknown)
def get_k(self):
return get_k(self.df, self.identifiant, self.unknown)
def get_k(self):
return get_k(self.df, self.identifiant)