Object-Centric Process Mining

RWTH Aachen University via edX

Go to class Write review

Details

Go to class

Provider

edX
Pricing

Free Online Course (Audit)
Languages

English
Certificate

$99.00 Certificate Available
Duration & workload

10 weeks, 3-6 hours a week
Sessions

On-Demand
Subtitles

English

Found in

Overview

This course covers state-of-the-art object-centric process mining methods and tools to enable participants to get a comprehensive understanding of the capabilities and use cases for object-centric process mining. The content covers how process mining can be used to understand a process, check its correctness, and apply machine learning methods to improve all types of processes.

Traditional process mining is often limited to analyzing processes centered on a single case identifier. Object-Centric Process Mining (OCPM) supports the analysis of processes involving multiple interacting objects (e.g., customers, orders, products, invoices) within a single model. As a result, data need to be extracted only once, distortions are avoided, and performance problems involving multiple processes or organizational units can be identified.

First, sources of event data are discussed. With the rise of digitalization, more and more events of every process are tracked digitally. Object-centric event logs store this data, which enables the computation of various process insights. After covering the most important process modeling notations (including state-of-the-art object-centric process model notations), process discovery approaches are presented. They can automatically learn a process model from event data. Then, the course describes conformance-checking methods that can identify behavioral differences between the desired process and the behavior observed in reality. The course also covers approaches and tools to analyze the performance and organizational structure of processes. Finally, the connection between process mining and machine learning is discussed, by describing how process mining can identify relevant problems in processes and transform them into machine learning problems.

Throughout the course, the concepts explained in the videos are accompanied by hands-on quizzes and optional coding and tool practices. These practical experiences foster a better understanding of algorithms and provide a guided introduction to state-of-the-art process mining tools.

After taking the course, students should have a great understanding of the different process mining techniques and should be comfortable applying them to object-centric event data to improve their processes.

Syllabus

Week 1: Introduction to Process Mining

In the first week of the course, we will give an overview of process mining and explain why an object-centric perspective on processes can improve process insight. We describe sources of event data and how to extract that type of data. Also, we introduce concepts like flattening (focusing on one process participant), which are commonly used throughout the course.

Week 2: Process Models

In the second week, we introduce the first process models. There are multiple process model notions that can describe the order of activities in a process. This week is all about directly follows graphs and Petri nets. We first introduce them for traditional (case-centric) process mining and then extend that to object-centric perspectives. Directly follows graphs are commonly used in industry and Petri nets have a strong theoretical background.

Week 3: Comparing Process Models

In the third week, we first introduce more process models, like BPMN and Process Trees and then discuss how the quality of process models can be compared to each other. We discuss the representational bias of the presented model notations and introduce concepts like confusion matrixes to compare quality criteria between models quantitatively.

Week 4: Process Discovery 1

In the fourth week, we will present the first process discovery algorithms. We start with a directly follows graph miner that is used to compute many of the directly follows graphs in commercial process mining tools. Then we cover the alpha miner, which is the first process discovery algorithm.

Week 5: Process Discovery 2

The fifth week continues with process discovery. Process discovery is an essential step in each process mining project to uncover the actual process behavior based on real event data. This week, we focus on more advanced approaches like region-based mining and inductive mining and extend the traditional case-centric process mining approaches to object-centric conformance checking.

Week 6: Conformance Checking

The sixth week focuses on process conformance checking. In conformance checking, one wants to detect behavior in the real process that deviates from the desired behavior described by a normative process model. We present three approaches: Causal footprints, token-based replay, and alignments.

Week 7: Beyond Control Flow

In the seventh week, we extend the view on processes and focus on other important parts of a process than the pure ordering of activities. We introduce dotted charts, organizational mining, social networks, and performance analysis. Then, we combine all these perspectives in an integrated process model.

Week 8: Link to Machine Learning

In week eight, we discuss the link from process mining to machine learning. Process mining serves as a lens on event data, allowing us to understand and identify concrete process problems. Using situation tables, these problems can then be transferred to supervised or unsupervised machine learning problems.

Week 9: Outlook and Closure

Week nine summarizes the previous weeks' content and gives a process mining outlook. We present and discuss more practical challenges regarding executing a process mining project and scaling process mining for large amounts of data. Also, digital twins and generative AI and their relation to process mining are discussed this week. The course closes with an outlook on future developments in process mining.