Institut für Operations und Information Systems
Mag.
Dr. Elmar Steiner
DOKTORAT
Institut für Operations und Information Systems
Betreuer:
Univ.-Prof. Dr.
Ulrich Pferschy
Optimization, Simulation and Machine Learning for Three
Applied Decision-Making Problems under Uncertainty
Many decisions in business and public administration must be made before all relevant information is available. This cumulative dissertation examines this fundamental problem in three application fields: timetabling in teacher education, the control of heuristic search algorithms, and the reprocessing of used products in the circular economy. What they share is that uncertainty is not merely an input but arises from the behaviour of those involved and from earlier decisions themselves. For timetabling at the University of Graz, a multi-stage integer optimization model is developed that produces timetables with few conflicts despite freely combinable subjects and individual study paths, and allocates students to courses fairly. In the second field, an agent uses reinforcement learning to balance a search heuristic between exploration and exploitation at runtime. In the third field, a learning agent is trained in a discrete-event simulation to decide on resale, refurbishment or disposal of returned products. It finds non-trivial strategies adapted to the cost structure. A modular simulation toolkit additionally gives small organizations easier access to such decision support. The thesis shows that only the combination of methods enables robust decisions: optimization ensures structure, simulation substitutes for missing data, and learning provides adaptivity. Open questions concern above all explainability, fairness and validation with real-world data.