def random_assign(self, event: Event, target: Event, p: float):
convert(event)
convert(target)
groupby_object = self.df.groupby(event.keys())
name = tuple(event.values())
if name.__len__() == 1:
name = name[0]
try:
index = groupby_object.get_group(name).index
target_index = random.sample(list(index),
int(round(index.__len__() * p)))
col = self.df.columns.get_loc(target.keys()[0])
self.df.ix[target_index, col] = target.values()[0]
except KeyError:
pass
def convert(event: Event):
for k in event.keys():
if isinstance(k, int):
name = event.name[k].name
v = event.dict[k]
event.dict.pop(k)
event.dict[name] = v
def count(self, event: Event, condition: Event, relationship: str):
convert(event)
convert(condition)
groupby_object = self.df.groupby(condition.keys())
name = tuple(condition.values())
if name.__len__() == 1:
name = name[0]
try:
group = groupby_object.get_group(name)
key0 = event.keys()[0]
key1 = event.keys()[1]
if relationship == 'equal':
sub_group = group[group[key0] == group[key1]]
return sub_group.__len__() / group.__len__()
else:
sub_group = group[group[key0] != group[key1]]
return sub_group.__len__() / group.__len__()
except:
return 0
def get_marginal_prob(self, event: Event):
convert(event)
groupby_object = self.df.groupby(event.keys())
name = tuple(event.values())
if name.__len__() == 1:
name = name[0]
try:
return groupby_object.get_group(name).__len__()
except:
return 0