In this talk (/workshop) we introduce the field of process mining as well as the PM4Py library, i.e., a python library implementing process mining algorithms. Process mining is a collection of tools, techniques, methods etc. aimed at analyzing operational process data, stored during the execution of (business) processes.
The execution of business processes, in modern organizations, is often supported by different information systems. Clearly, the better one understands its core processes, the easier to steer the process towards an increased overall process performance. In order to gain a detailed understanding of the execution of a process, i.e., in terms of its performance and conformance, different techniques and methods originating from the process mining domain are typically used. Essentially, process mining techniques aim to distill actionable knowledge and insights of a process, on the basis of historical execution data. For example, process discovery algorithms are able to translate the captured even data into a process model, e.g., into a BPMN model. Furthermore, conformance checking algorithms allow us to compute, in an exact manner, whether or not the execution of the process, i.e., as captured in the data, is in line with a reference model. Additionally, heaps of techniques exist that allow us to compute insights in the performance of the process.
Over the past twenty years, process mining has remained a relatively unknown, scientific, endeavor. However, recently, there has been a keen interest from industry in this technology as well. This is best illustrated by the German process mining company Celonis, recently valued over 1 billion dollars of net worth. Furthermore, the 1st international conference on process mining (https://icpmconference.org), held in June 2019, attracted a total number of 20 sponsors from industry (mainly process mining vendors) and a total of 400 participants.
Despite the vast increase in commercial/industrial interest, the amount of open-source tools supporting process mining techniques has, until recently, been very little. For years, ProM (http://promtools.org) and Apromore (https://apromore.org/) have been the de-facto leading open-source process mining solutions. However, these tools are primarily designed to be used by end-users, i.e., they provide a front-end and are not easily integrated in other software solutions. Only recently, solutions in R, i.e., bupaR (https://www.bupar.net/) and Python, i.e., PM4Py (https://pm4py.fit.fraunhofer.de), have been developed.
In this talk, we present the main idea of process mining, i.e., going from captured event data process insights, and, we briefly show an example in python. In case of a possible workshop, we will elaborate more on the algorithmic details of one (of the many) state-of-the-art process mining algorithms.