Metaheuristic nurse scheduling with hospital clustering using flower pollination algorithm

Abstract

Effective nurse scheduling is essential to ensure balanced workloads, reduce fatigue, and maintain healthcare service quality. However, the nurse scheduling problem (NSP) is complex due to constraints related to nurse skills, task requirements, and legal working-hour limits. This study proposes an integrated framework combining a mathematical optimization model with metaheuristic algorithms to generate optimal daily nurse activity schedules. Genetic algorithm (GA) and simulated annealing (SA) are employed to produce near-optimal solutions for nurse populations ranging from 3 to 50 individuals, considering skill-level compatibility, workload balance, and maximum working hours. Experimental results using real scheduling data from 30 nurses across three skill levels demonstrate that all generated schedules satisfy the imposed constraints, with no nurse exceeding the 12-hour daily working limit. Comparative analysis shows that GA achieves lower scheduling costs for larger nurse populations, while SA consistently requires significantly shorter computation times, making it suitable for time-sensitive applications. In addition, the flower pollination algorithm (FPA) is used to cluster 3,155 hospitals based on bed capacity, service variety, and workforce size, supporting data-driven workforce distribution analysis. The proposed framework integrates operational scheduling optimization with hospital-level clustering, providing practical decision support for healthcare workforce planning.