Benchmarking the Nurse Re-Rostering Problem: A Dataset and Instance Generator for evaluating Algorithms that handle short-term scheduling Disruptions

Abstract

The Nurse Re-Rostering Problem (NRRP) involves short-term schedule adjustments in response to disruptions such as staff absences or sudden changes in demand. Unlike the well-studied Nurse Rostering Problem, the NRRP receive relatively little attention-both in terms of algorithm development and in terms of the availability of standardized benchmark datasets. Existing benchmarks primarily focus on initial rostering. To address this gap, we propose a configurable benchmark instance generator specifically designed for the NRRP, along with two benchmark datasets covering planning horizons of 14 and 28 days and varying levels of complexity. The generator enables researchers to simulate realistic disruption scenarios with adjustable complexity, while the datasets support reproducible and standardized evaluation of re-rostering algorithms. This benchmark dataset facilitates fair comparisons and drives progress in algorithmic research for the NRRP.