Theses
On this page, you can find all necessary information for writing a thesis with us.
Process
If you are interested in one of the offered thesis topics, please proceed as follows:
Send us a one-page cover letter, in which you outline your
- Interests
- Experiences (e.g., completed courses, projects, work experience)
- Motivation to contribute these interests and experiences to the chair
- and potentially relevant publications.
Additionally, it is advantageous if you have already taken courses in the field and have experience with academic writing and research – please make this clear in your cover letter. Send the documents to thesis(at)infs.cit.tum.de.
If your cover letter is informative and convincing, we will have an initial meeting to specify the thesis topic. Following this, you will complete an exposé within 4 – 6 weeks. In it, you will provide
- Motivation (research question and relevant problems/research gaps, supported by references from academic literature)
- Planned objectives/contributions (specific milestones and objectives)
- Methodology and Evaluation (planned methods to answer research questions and evaluate results)
- Risks (potential risks that may hinder the progress and success of your work).
We may request changes to the exposé, so it may go through several iterations if necessary.
Once the exposé forms a satisfactory basis for the thesis, you can register the work. We recommend conducting regular meetings with the supervisors during the working phase to address questions and uncertainties directly, ensuring a satisfactory outcome for all parties involved.
The deadline is 4 months later for the Bachelor's programs in Informatics and Information Engineering. If you are enrolled in a different program, please consult the corresponding FPSO. Typically, depending on the FPSO, the Bachelor colloquium takes place after the end of the processing period.
For formalities regarding registration, submission, etc., consult your FPSO and visit the official pages of TUM, e.g.https://www.cit.tum.de/en/cit/studies/students/thesis-completing-your-studies/informatics/. We would also like to explicitly point out the rules for good academic practice which you have to consider and which you can find at https://portal.mytum.de/archiv/kompendium_rechtsangelegenheiten/sonstiges/TUM_SGwP_en.pdf/.
Topics
Below you will find the subject areas and specific problem statements for which you can write Bachelor's or Master's theses with our department.
Topic Areas
General topics for which we conduct thesis work include:
- Intelligent Process Automation / Hyperautomation
- Business Process Simulation (BPS)
- Sustainability in BPM
- Resource Allocation in Business Processes
- Explaining Activity Dependencies.
Concrete Topic Suggestions
Bachelor Thesis Topics
To test and experiment with ideas and newly developed approaches in the BPM area, we need real-use case scenarios that illustrate challenges and allow the testing of developed techniques. This thesis focuses on documenting a business process, with the domain being freely chosen. In addition to creating a process model, the focus is on defining dependencies between activities and the contextual influence on the process. Furthermore, an extended event log will be generated, e.g., through simulation of the process model, containing attributes that capture the contextual information.
Objective/Task: Provide a use case where all activity dependencies are specified and the contextual influence is clearly defined. A process model should be designed, and an event log generated.
Traditional vendors rely on intermediate distributors to sell and ship their products to customers, storing their goods in warehouses and using traditional shipping routes In contrast to this, emerging online retailers (such as Temu or Aliexpress) allow shipping goods directly to customers, usually via airplane to shorten delivery times. However, one major risk is that customs withholds or rejects these goods, either due to faulty customs declarations or missing certificates. The major question to be addressed in this thesis is in how far the environmental impact of the two types of supply process differs.
Goal: Model the two supply process variants; analyse and compare their environmental impact using life cycle assessment (LCA) and the SOPA framework for sustainability-oriented process analysis. Make justified assumptions, where necessary
During process execution, upcoming process tasks must be allocated to internal resources, such as humans or machines. The allocation of tasks to resources is a complex decision-making problem with a high impact on the effectiveness and efficiency of business processes. A wide range of system-initiated (primarily automated) resource allocation approaches have been developed over the last decades. We have analyzed them in a structured literature review. In this, we have learned that benchmarking these approaches is still challenging. To prepare a benchmark, the identified studies of this literature review need to be further coded regarding different aspects, such as allocation goal, control-flow constraints (considered yes, no, assumptions), constraints regarding the resources, constraints regarding tasks, allocation capability, run-time/preplanning, etc.
Goal: Develop a data set and a search interface to identify resource allocation studies for a benchmark.
During process execution, upcoming process tasks must be allocated to internal resources, such as humans or machines. Allocating tasks to resources is a complex decision-making problem that has a high impact on the effectiveness and efficiency of business processes. Reinforcement learning is a promising concept to support the allocation of tasks in business processes. Multiple studies have been proposed in this field that show similarities and differences in modeling the states of the environment and the action of the agent.
Goal: Identify and analyze research works using reinforcement learning and identify similarities and differences. Propose a more general approach for reinforcement learning in the resource allocation for business processes and define configuration possibilities.ign and automation of business processes.
Organizations carry out a variety of business processes to meet the needs of their customers. These processes, which consist of coordinated activities designed to accomplish specific objectives, are collectively known as business processes. Business Process Simulation (BPS) plays a crucial role in this context by providing a powerful tool to create virtual environments for testing process modifications without actual real-world consequences. Machine learning is one of the recent technologies that are used to enhance and optimize business process simulation. Goal: to provide an overview of recent efforts in integrating machine learning technologies to enhance business process simulation.
Understanding the contextual rationale behind activity relationships in business process models is crucial for preventive business process redesign. Based on the motivational origin of an activity relationship, we can derive risks and consequences when changing it. Previous approaches have explored the extraction of explanatory rationales using Large Language Models (LLMs) and questionnaires. However, these methods rely on post-hoc analysis rather than integrating rationale elicitation into the modeling process itself. This thesis topic aims to develop an interactive approach that captures the rationale for activity relationships during business process modeling. Specifically, when a relationship is added, the modeler will be prompted to specify whether it is driven by legal requirements, business rules, or best practices. To achieve this, a user interface will be implemented, for example, based on BPMN-js, seamlessly integrating rationale elicitation into the modeling workflow.
Objective/Task: Develop a prototypical UI so that users have to provide an explanatory rationale for each relationship they are modeling. The entered data should be stored accordingly.
Bachelor or Master Thesis Topics
Business process simulation is a key method for predictive process improvement. The more precise a simulation model reflects (potential) reality, the more valuable insights can be derived from it. To this end, information from multiple sources, such as process mining tools, databases, and human knowledge must be considered. This presents a challenge to simulation experts, who then need to manually integrate different tools and their data formats. Visual models promise to make simulation model discovery pipelines tangible and automatable, addressing this challenge.
Goal: Develop a user-friendly mechanism to generalize the integration of different tools for the discovery and parameterization of business process simulation models. Evaluate the developed artifact.
Business process management assumes a clearly defined start and end of business processes. However, not all business processes adhere to this logic, and as a consequence, BPM analysis techniques cannot appropriately capture their behavior. In previous work, we introduced the notion of vitalizing business processes that target the lifecycle process of one or more entities. The MedEvaluate the Concept of Vitalizing Business Processes on the MIMIC-IV data setical Information Mart for Intensive Care (MIMIC)-IV database is a publicly available relational database that includes data on patient treatment in a tertiary academic medical center in Boston, USA. It provides complex care processes in a hospital from end to end. It includes many of the vitalizing business processes.
Goal: Apply the notation of vitalizing business processes and the proposed concepts to one patient group of the MIMIC-IV data set. Evaluate the completeness and usefulness of the proposed concepts.
Organizations execute diverse business processes to serve their customers. Business processes can be defined as a series of structured, coordinated activities aimed at achieving goals (e.g., hiring a new employee). Business Process Simulation (BPS) steps in here – a potent tool creating a virtual environment for testing process changes without real-world effects. The quality of input data is crucial for the effectiveness of business process simulation, particularly event logs, which capture real-life process executions. However, real-world event logs often suffer from various quality issues, affecting simulations' accuracy and reliability.
Goal: aims to provide a comprehensive overview of data quality challenges, specifically focusing on issues that directly impact simulation outcomes. then conducting extensive practical testing on available datasets to analyze their suitability and quality for simulation purposes.
This research intends to examine the current state of data and analytics architecture within the context of Enterprise Architecture Management (EAM). As organizations increasingly adopt data-driven decision-making, integrating data and analytics into EAM frameworks is crucial for ensuring strategic alignment, scalability, and governance. This study aims to identify and analyze existing literature to determine the current state in managing data and analytics architecture within EAM. By conducting a systematic literature review the research should provide insights into how organizations can optimize their enterprise architecture to better support data-driven strategies. The findings will contribute to a deeper understanding of the current state and summarize data and analytics capabilities within EAM frameworks.
Goal: Conduct a literature review to explore the current state of data & analytics architecture within EAM. Identify and summarize the key findings of the literature reviews and give recommendations.
Master Thesis Topics
Clinical studies are well documented in central storage, especially in the USA (https://clinicaltrials.gov/data-api/about-api). This availability of public data on clinical studies allows process mining to be applied to this data to find new insights or to evaluate new process mining techniques on this data set. However, in preparation, the data must first be transferred into an event log. Therefore, decisions on the case notion, the selected event types, and their attributes have to be made. Experts in this field can support the process of extracting an event log.
Goal: Follow the procedure of building a useful event log by Jans et al. and extract an event log with the help of experts with whom we are in contact for applying process mining techniques on the publicly available clinical study data.
Knowledge is a comprehensive term but essential for various research domains. Also, in Business Process Management (BPM), process knowledge is crucial for understanding processes and dependencies between individual activities. Knowledge includes more than just execution sequences and data encoded in event logs. To make the concept of process knowledge more tangible, the theory of knowledge and knowledge management should therefore be examined more closely in this thesis. What is process knowledge? How can we classify this kind of knowledge? Are there different types of knowledge that are represented and encoded differently? How can we extract knowledge from data and use this knowledge, for example, for process redesign?
Goal: Conduct a literature review on the theory of knowledge and relate the concepts with process knowledge to support the redesign and automation of business processes.
Due to the constantly changing environment, business processes must adapt their process behavior. This means that relations between the activities of a process are changed to make it more efficient with regard to cost, time, and quality. Business process simulation helps in testing new process scenarios safely in a computer-imitated environment and can measure the implications on performance, quality, and costs. However, not every relationship between activities can be changed. Some of them are essential for proper process termination and has to be ensured. Other relationships can be changed but are associated with certain risks and consequences. This thesis is therefore concerned with how business process simulation can be further developed for process behavior redesign so that the vulnerability of relations is taken into account. Specifically, it will be investigated to what extent machine learning approaches, in particular, can be used to develop new process scenarios that take into account the vulnerability of activity relation learnt from historical data and provided by human input.
Goal: Development of a machine learning approach which identifys and tests new process scenarios with a process simulation changing the process behavior by taking into account the vulnerability of activities.
Conformance checking is a fundamental technique in process mining, used to assess whether a business process model accurately reflects real-world process execution recorded in event logs. Traditionally, conformance checking compares a process model against execution traces to identify deviations. This research introduces an alternative approach by leveraging activity relationship types instead of event logs to evaluate conformance. The approach is based on the activity relationships matrix, which defines relationships between activity pairs using two types of dependencies: temporal dependency and existential dependency. This allows for a structured representation of how activities relate to a process. The goal of this research is to develop an algorithmic method to compare a given process model with the activity relationships matrix and determine to what extent the discovered relationships conform to the predefined matrix. A prototype implementation will be developed to validate the approach, providing insights for identifying deviations based on relationships. This research aims to enhance the precision and flexibility of conformance checking by shifting the focus from event-driven evaluation to structural relationship validation.
Ongoing and Past Theses
You can find currently ongoing and past theses here.