Martin, Jan Niklas and Wilsdorf, Pia and Uhrmacher, Adelinde M. (2026) Calibrating Simulation Models using only Domain Knowledge. In: 2026 Winter Simulation Conference, 6-9 Dec 2026, Glasgow, Scotland, UK.
Full text not available from this repository.Abstract
When suitable empirical data for validating simulation models are unavailable, stylized facts provide a valuable alternative. Despite their usefulness, two challenges limit their broader adoption: the lack of concise, unambiguous, and reproducible specifications that allow automatic checking, and the lack of support for their use beyond binary tests for model validation. We present SF-CalSim, which combines a) a domain-specific language for succinctly specifying stylized facts, b) a quantitative metric for measuring the degree of satisfaction, and c) Swarm Optimization and Approximate Bayesian Computation for estimating parameter values and the posterior distribution, respectively. We apply the approach to the German Epidemic Microsimulation System using independently defined stylized facts, we recover parameterizations close to the original model and demonstrate their robustness.
| Item Type: | Conference or Workshop Item (Paper) |
|---|---|
| Additional Information: | accepted |
| Projects: | E-MoSi-CA |
