Understanding and Supporting Software Model Evolution through Edit Operation Mining and AI-based Software Model Completion

Description

Model-based Systems Engineering has become increasingly important in managing the complexity of modern software systems. However, the evolution of software models in large-scale, real-world projects remains a significant challenge due to the lack of effective and automated methods. This thesis addresses this problem by providing a comprehensive theoretical foundation for software model evolution, and, based on this theory, presents practical approaches to understand and support model evolution, which will be evaluated and validated by empirical studies. The first part of the thesis develops a theory to software model evolution and introduces novel techniques—backing the theory—based on graph mining, large language models, and graph neural networks. These methods allow to automatically define model transformations, and support software model completion, solely based on model histories from model version control systems. The second part motivates and shows the relevance and usefulness of the research behind the thesis by providing insights into the complexity of a real-world industrial model-driven product line at our industry partner. In the third part, controlled experiments provide evidence for the theory developed in the first part of the thesis. The findings demonstrate the feasibility and that there is a potential for data-driven intelligent modeling assistants to support model-based engineering—not only boosting productivity of engineers, but also accuracy and overall quality of software models. This research contributes to advancing Model-based Systems Engineering by proposing automated solutions that support the continuous evolution of software models, ultimately leading to more robust and maintainable systems.

Period	24 Oct 2025
Work for	Universtität Saarbrücken, Germany
Degree of Recognition	National