def dataframe(self):
"""
Return pandas DataFrame containing the event log in matrix format.
:return: pandas.DataFrame
"""
start_event = Event(timestamp=None,
**dict((a, EventLog.start_symbol)
for a in self.event_attribute_keys))
end_event = Event(timestamp=None,
**dict((a, EventLog.end_symbol)
for a in self.event_attribute_keys))
frames = []
for case_id, case in enumerate(self.cases):
if case.id is not None:
case_id = case.id
for event_pos, event in enumerate([start_event] + case.events +
[end_event]):
frames.append({
'case_id':
case_id,
'event_position':
event_pos,
'name':
event.name,
'timestamp':
event.timestamp,
**dict([
i for i in event.attributes.items() if not i[0].startswith('_')
])
})
def generate_random_event(self):
if self.activities is None:
raise RuntimeError('activities has not bee set.')
event = Event(name=f'Random activity {np.random.randint(1, len(self.activities))}')
if self.attributes is not None:
event.attributes = dict(
(a.name, f'Random {a.name} {np.random.randint(1, len(a.values))}') for a in self.attributes)
return event
def generate_random_event_v2(self, act):
if self.activities is None:
raise RuntimeError('activities has not bee set.')
actset = [x for x in self.activities if x not in act]
event = Event(name=f'{np.random.choice(actset)}')
if self.attributes is not None:
event.attributes = dict(
(a.name, f'Random {a.name} {np.random.randint(1, 3)}')
for a in self.attributes)
return event
def from_sql(server, database, resource, password, schema='pm'):
import pyodbc
conn = pyodbc.connect(f'DRIVER={{ODBC Driver 17 for SQL Server}};'
f'SERVER={{{server}}};'
f'DATABASE={{{database}}};'
f'UID={{{resource}}};'
f'PWD={{{password}}}')
cursor = conn.cursor()
cursor.execute(
f'SELECT CaseId, ActivityName, Resource, Timestamp FROM {schema}.EventLog ORDER BY CaseId, Timestamp, SortKey, ActivityName'
)
event_log = EventLog()
case = None
current_case_id = None
for row in cursor.fetchall():
case_id = row[0]
timestamp = '' if row[3] is None else row[3].strftime(
'"%Y-%m-%d %H:%M:%S"')
name = row[1]
resource = row[2]
if case_id != current_case_id:
case = Case(id=case_id)
event_log.add_case(case)
current_case_id = case_id
case.add_event(
Event(name=name, timestamp=timestamp, resource=resource))
def from_csv(file_path):
"""
Load an event log from a CSV file
:param file_path: path to CSV file
:return: EventLog object
"""
# parse file as pandas dataframe
df = pd.read_csv(file_path)
# create event log
event_log = EventLog()
# iterate by distinct case id
for case_id in np.unique(df['case_id']):
_case = Case(id=case_id)
# iterate over rows per case
for index, row in df[df.case_id == case_id].iterrows():
start_time = row['start_time']
end_time = row['end_time']
event_name = row['event']
user = row['user']
_event = Event(name=event_name,
timestamp=start_time,
end_time=end_time,
user=user)
_case.add_event(_event)
event_log.add_case(_case)
return event_log
class ReworkAnomaly(Anomaly):
"""Insert 1 sequence of n events coming from the case later in the case."""
def __init__(self, max_sequence_size=2, max_distance=0):
self.max_sequence_size = max_sequence_size
self.max_distance = max_distance
super(ReworkAnomaly, self).__init__()
def apply_to_case(self, case):
if len(case) <= 1:
return NoneAnomaly().apply_to_case(case)
size = np.random.randint(2, min(len(case), self.max_sequence_size) + 1)
start = np.random.randint(0, len(case) - size + 1)
distance = np.random.randint(
0,
min(len(case) - (start + size), self.max_distance) + 1)
t = case.events
dupe_sequence = [Event.clone(e) for e in t[start:start + size]]
inserted = [e.json for e in dupe_sequence]
anomalous_trace = t[:start + size +
distance] + dupe_sequence + t[start + size +
distance:]
case.events = anomalous_trace
case.attributes['label'] = dict(anomaly=str(self),
attr=dict(size=int(size),
start=int(start + size +
distance),
inserted=inserted))
return case
def from_dict(log):
event_log = EventLog(**log['attributes'])
for case in log['cases']:
_case = Case(id=case['id'], **case['attributes'])
for e in case['events']:
event = Event(name=e['name'],
timestamp=e['timestamp'],
**e['attributes'])
_case.add_event(event)
event_log.add_case(_case)
return event_log
def path_to_case(self, p, label=None):
"""
Converts a given path to a case by traversing the graph and returning a case.
:param p: path of node identifiers
:param label: is used to label the case
:return: a case
"""
g = self.graph
case = Case(label=label)
for i in range(0, len(p), self.event_len):
event = Event(name=g.node[p[i]]['value'])
for j in range(1, self.event_len):
att = g.node[p[i + j]]['name']
value = g.node[p[i + j]]['value']
event.attributes[att] = value
case.add_event(event)
return case
def parse_event(event):
attr = dict(
(attr.attrib['key'], attr.attrib['value']) for attr in event)
timestamp = None
# if 'time:timestamp' in global_attr['event'].keys():
if 'time:timestamp' in attr:
timestamp = attr['time:timestamp']
name = ''
if len(classifiers) > 0:
keys = classifiers[0]['keys']
check_keys = [key for key in keys if key not in attr]
if len(check_keys) > 0:
print(
f'Classifier key(s) {", ".join(check_keys)} could not be found in event.'
)
return None
values = [attr[key] for key in keys]
name = '+'.join(values)
return Event(name=name, timestamp=timestamp, **attr)
def from_json(file_path):
"""
Parse event log from JSON.
JSON can be gzipped
:param file_path: path to json file
:return:
"""
if not isinstance(file_path, str):
file_path = str(file_path)
if file_path.endswith('gz'):
import gzip
open = gzip.open
# Read the file
with open(file_path, 'rb') as f:
log = json.loads(f.read().decode('utf-8'))
event_log = EventLog(**log['attributes'])
# Compatibility: Check for traces in log
if 'traces' in log:
case_key = 'traces'
else:
case_key = 'cases'
for case in log[case_key]:
_case = Case(id=case['id'], **case['attributes'])
for e in case['events']:
event = Event(name=e['name'],
timestamp=e['timestamp'],
**e['attributes'])
_case.add_event(event)
event_log.add_case(_case)
return event_log
def clone(trace):
events = [Event.clone(event) for event in trace.events]
return Case(id=trace.id, events=events, **dict(trace.attributes))