Hardness studies for nurse rostering problems

We present an empirical hardness model for nurse rostering by explicitly building on previous developed models for SAT problems. The resulting model allows to predict algorithm performance based on features of the problem instances in a specific distribution for specific performance criteria. In doing so we demonstrate that the performance of an algorithm on an instance of a real world problem can be estimated from a limited number of indicators. We speculate that a similar approach may be valid in other application domains.status: publishe