Authors:
Fang Wan
1
;
Julien Fondrevelle
1
;
Tao Wang
2
;
Kezhi Wang
3
and
Antoine Duclos
4
Affiliations:
1
INSA LYON, Université Lyon2, Université Claude Bernard Lyon1, Université Jean Monnet Saint-Etienne, DISP UR4570, Villeurbanne, France
;
2
Université Jean Monnet Saint-Etienne, INSA LYON, Université Lyon2, Université Claude Bernard Lyon, DISP UR4570, Roanne, France
;
3
Department of Computer Science, Brunel University London, Uxbridge, Middlesex, U.K.
;
4
Research on Healthcare Performance RESHAPE, INSERM U1290, Université Claude Bernard, Lyon 1, France
Keyword(s):
Surgery Scheduling, Large Language Model, Combinatorial Optimization, Multi-Objective.
Abstract:
Large Language Model (LLM) have recently been widely used in various fields. In this work, we apply LLMs for the first time to a classic combinatorial optimization problem—surgery scheduling—while considering multiple objectives. Traditional multi-objective algorithms, such as the Non-Dominated Sorting Genetic Algorithm II (NSGA-II), usually require domain expertise to carefully design operators to achieve satisfactory performance. In this work, we first design prompts to enable LLM to directly solve small-scale surgery scheduling problems. As the scale increases, we introduce an innovative method combining LLM with NSGA-II (LLM-NSGA), where LLM act as evolutionary optimizers to perform selection, crossover, and mutation operations instead of the conventional NSGA-II mechanisms. The results show that when the number of cases is up to 40, LLM can directly obtain high-quality solutions based on prompts. As the number of cases increases, LLM-NSGA can find better solutions than NSGA-II.