Predicting employee absenteeism to generate robust rosters

Employee absences often lead to disruptions in rosters, necessitating last-minute changes to employee schedules. A common strategy to minimize the adverse effects of these changes is to assign employees to on-call duties, thereby increasing the robustness in the rosters. This study explores the effectiveness of a data-driven robust rostering approach, using predictions of employee absences to schedule an appropriate number of on-call duties. Numerical experiments demonstrate how the accuracy of absence predictions significantly impacts the robustness of the resulting rosters. We introduce a methodology to assess the conditions under which a data-driven robust rostering approach can outperform simple, non-data-driven rostering strategies.