Fundamentals of Process Mining

URL study guide

Omschrijving

This course is exclusively available to students enrolled in the Bachelor programs in Computer Science and Data Science.

In many organizations, work is typically organized in the form of processes. Requests, orders, offers, invoices, or in general ‘cases’, are all handled by organizations for their customers or clients according to pre-defined activities executed by their staff. Event data records multi-variate event sequences of who executed which activity, at which point in time, for which case along, with additional attributes relevant to the process case. The sequence of events recorded for a case reflects the specific activities performed for this case. Such event data is omnipresent and can be used to gain insights into the efficiency of an organization as well as into the compliance of processes with quality requirements and regulations. However, using traditional machine learning or data mining techniques on sequential event data is not trivial, as many of these techniques are not designed to work with end-to-end processes.

This course introduces Process Mining as an emerging discipline that seeks to analyze event data by “discovering” process models that help describe, evaluate and monitor end-to-end processes. The course covers the following broad topics:

- Introduction: We begin by introducing core concepts such as processes, process models, and event data. As part of this introduction, we learn how to read and create basic process models, which will give us a basis to understand process behavior later on.
- Event Log Exploration and Descriptive Analytics: We then move to analyzing event logs. We first engage in event log exploration through tasks such as case and event filtering and assessing data quality. We then learn how to conduct variant analysis to understand and compare different executions of the same process, and how to define and measure process performance across multiple dimensions. Using descriptive analytics, we analyze and visualize both process variability and performance.
- Process Discovery with Directly-Follows Graphs: Next, we focus on process discovery. We learn how to extract a simple type of process model, called a directly-follows graph (DFG), from an event log using basic techniques available in many commercial tools. We analyze DFGs from different angles and learn what their limitations are.
- Prediction and Rule Induction: We then introduce predictive process monitoring and rule induction. We focus on tasks such as outcome prediction and next-step prediction, and learn how to do feature engineering to prepare event logs for machine learning techniques. We introduce rule induction as a way to derive simple, interpretable rules that help explain process behavior.
- Putting Everything Together: Finally, we discuss how to run a process mining project in practice. We will learn about the typical steps and general guidelines for applying process mining in real-world settings, from data preparation to communicating findings.

Throughout the course, we use existing process mining tools such as Fluxicon Disco (through an academic license) and the Python framework PM4Py. We also work with real-life event logs to gain hands-on experience analyzing complex data.

Doelstellingen

Beoordelingsmethode