Bachelor Thesis - Predictive Process Monitoring



M. Seran Uysal

Scientific Assistant


+49 241 80 21904




Execution management and therefore operational excellence requires to predict and forecast unwanted behavior before it arises. Ultimately, customers are looking for a prescriptive prediction which is understandable to humans and can be automated.

Customers are interested in questions like:

  • Will this prospective customer finally sign? What needs to be done next?
  • How long will it take to successfully resolve this service ticket?
  • How to prevent costs from missing SLAs?

Celonis is heavily investing in the predictive process monitoring journey and is therefore interested in building up knowledge and experience in this field.

To get started, a thesis candidate conducts literature research and outlines state of the art approaches including advantages and disadvantages in terms of accuracy, complexity but also run time behavior. The focus for first implementations is on time and its variations, like next occurrence of an activity and remaining time of a running case. The thesis candidate is not restricted to pure statistical prediction and regression algorithms but could also explore in the area of neural networks and LSTMs. A candidate should explore different variations of encodings, bucketings and prediction algorithms. The achieved results are compared to publications as well as to comparable results generated by other tools.

If the first part is successfully completed and time allows for a continuation, a candidate would also contribute scientifically. Recently, the idea of object-centric process mining was proposed which is a paradigm shift and influences how process mining algorithms are working. The idea is to research the influence of object-centric event logs on time related predictions if multiple objects with shared activities are included. Based on some ideas outlined here, the predictions would be influenced by shared activities which act as synchronization points.


  • Bachelor student of computer science or a comparable program
  • Experience with Python, Java or even C++
  • Basic experience with statistical learning algorithms or even machine learning algorithms
  • Interest in developing research prototypes in order to conduct large-scale experiments
  • Strong communication skills and interest in presenting and testing your ideas
  • Good knowledge of spoken and written English
  • At least 3-6 months of part-time availability


Supervisor Wil van der Aalst


Andreas Swoboda

Nikou Günnemann

For more Information

Send an e-mail to for the application. Make sure to include all detailed information about your background and all scores for completed courses, and write “Celonis Bachelor Thesis: Predictive Process Monitoring" in the email's subject.