EURO 2024 Copenhagen
Abstract Submission

EURO-Online login

2949. Using predictions on employee absenteeism to generate robust personnel rosters

Contributed abstract in session MA-58: Automated Timetabling, stream Automated Timetabling.

Monday, 8:30-10:00
Room: S07 (building: 101)

Authors (first author is the speaker)

1. Pieter Smet
Computer Science, KU Leuven
2. Martina Doneda
DEIB, Politecnico di Milano
3. Giuliana Carello
Elettronica, Informazione e Bioingegneria, Politecnico di Milano
4. Ettore Lanzarone
DIGIP, University of Bergamo
5. Greet Vanden Berghe
Computer Science, KU Leuven

Abstract

Employee absenteeism occurs when an employee is not present at work during their scheduled hours. When this situation occurs, the absent employee's shift must be covered by another employee. This in turn may affect others in the organisation, creating an undesirable ripple effect that can propagate throughout the entire roster. Generating robust personnel rosters can proactively mitigate the negative consequences when employees are absent from work. One way of increasing roster robustness is by scheduling special on-call duties that can be converted into regular working duties whenever necessary. Machine learning models that predict when employee absences will occur provide an intuitive and regularly employed approach to determine where best to position on-call duties in a roster. In this talk we will propose a methodology to evaluate the circumstances under which such predictions can actually increase roster robustness. More specifically, we will analyze the results of a series of computational experiments to determine the prediction performance levels needed by a machine learning model to outperform a non-data-driven robust rostering method.

Keywords

Status: accepted

Back to the list of papers