procon is an extension to the popular pm4py process mining library with the goal to enable conformance checking on BPMN models. Assume you have an advanced BPMN 2.0 model, maybe with cancellation features, and you have gathered event data, e.g., from your ERP or CRM system. You are very likely interested to see how well the real worl process behavior matches the desired process model. procon gives you the opportunity to do that with only a few lines of Python code!
procon can be installed on Python 3.7.x / 3.8.x / 3.9.x by doing:
pip install -U proconThe following example shows you how to import a bpmn model and an event log to then derive conformance statistics based on them.
import os
import procon
# IMPORTANT: In case you want to make use of parallel programming, you need to encapsulate the conformance checking
# code inside a __name__ == '__main__' guard
if __name__ == '__main__':
# import .bpmn file
bpmn_graph = procon.import_bpmn(os.path.join("test_data","example.bpmn"))
# import event log as a dataframe
df = procon.import_event_log(os.path.join("test_data","example.xes"))
# alternatively, you can directly import a dataframe via pandas
# for more info see https://pm4py.fit.fraunhofer.de/documentation#importing
parameters = {}
# should boundary events be treated as labelled activities?
parameters['include_events'] = True
# derive alignemnts between event log and model
alignments = procon.compute_alignments(df, bpmn_graph, parameters=parameters)
# ideally, you save the alignments object in some file via pickle in case you do not want to wait the whole time again ;-)
# import pickle
# file_pi = open(os.path.join("path", "to", "alignments.obj"), 'wb')
# pickle.dump(alignments, file_pi)
# finally, derive conformance statistics from the alignments
res = procon.derive_statistics(alignments, df, bpmn_graph, parameters=parameters)
# save the resulting dataframe to a csv file on your machine
res.to_csv(os.path.join("conformance-result.csv"))Input BPMN model:
Output Table:
The meaning of the columns is as follows:
- Index/Activity: Name of the activity
- Occurrences: Total occurrences of the activity in the event log
Correct: Number of times the activity appeared at the desired point in time according to the process modelWrong Position: Number of times the activity appeared in a case in an undesired orderMissing: Number of times the activity was completely missing in a case although it was expected to appear