1738. A Multi-Neighborhood Simulated Annealing Approach for Solving the Integrated Healthcare Timetabling Problem
Invited abstract in session MD-10: Integrated Healthcare Timetabling Competition III, stream Automated Timetabling.
Monday, 14:30-16:00Room: Clarendon SR 1.06
Authors (first author is the speaker)
| 1. | Eugenia Zanazzo
|
| Polytechnic Department of Engineering and Architecture, University of Udine | |
| 2. | Sara Ceschia
|
| Polytechnic Department of Engineering and Architecture, University of Udine | |
| 3. | Roberto Maria Rosati
|
| WU Vienna University of Economics and Business | |
| 4. | Andrea Schaerf
|
| Polytechnic Department of Engineering and Architecture, University of Udine | |
| 5. | Pieter Smet
|
| Computer Science, KU Leuven | |
| 6. | Greet Vanden Berghe
|
| Computer Science, KU Leuven |
Abstract
In this talk, we present our Multi-Neighborhood Simulated Annealing approach to solving the newly proposed Integrated Healthcare Timetabling Problem.
The solver is designed in a fully integrated manner, meaning that at any time during the search, any aspect of a patient's stay or nurse assignment can be modified.
The search space is structured to exclude infeasible patient assignments while allowing the violation of certain hard constraints to improve connectivity and facilitate traversal. To efficiently explore this space, we defined five partially overlapping neighborhoods: four operate on variables related to different aspects of a patient's stay, including their admission schedule, assigned operating theater, and recovery room, while the fifth focuses on nurse assignments.
The search process begins from a randomly generated solution and progresses through a series of random moves. These moves are selected by first selecting the neighborhood based on pre-tuned probabilities and then randomly choosing a move within that neighborhood. The selected move is evaluated and executed if it satisfies the SA acceptance criterion.
To ensure a meaningful comparison between our results and those reported by the IHTC-2024 finalists, we set a 600-second time limit for the solver and performed parameter tuning exclusively on the publicly available dataset. Our results closely align with those of the IHTC-2024 finalists, demonstrating the effectiveness of our approach.
Keywords
- Health Care
- Timetabling
- Metaheuristics
Status: accepted
Back to the list of papers