Optimizing nurse scheduling problem using the hybrid of genetic algorithm and firefly algorithm

The Nurse Scheduling Problem is a type of scheduling problem aimed at fairly distributing shifts among hospital nurses while adhering to hard and soft constraints. Equitable workload distribution and balancing desirable/undesirable shifts are critical challenges in nursing management. This study proposes a solution for nurse scheduling that ensures fair shift allocation while incorporating nurses' preferences.

First, the research dataset was created using shift requests submitted by nurses, taking into account their preferences and the schedules provided by supervisors for 20 nurses over a 6-month period. One bit was considered for each shift for four states. Therefore, by including the shift schedule resulting from the proposed hybrid method, three datasets of 14, 880 bits each were formed. The hybrid model was configured, trained, and tested using this dataset. Next, the nurse schedules for the same 6-month period were generated using the trained model, compared with the supervisor's schedules. A paired T-test was conducted to evaluate the statistical significance of differences in three metrics: average night shifts, shift distribution balance, and nurse satisfaction.

The P values of the statistical tests for the three evaluation metrics: average night shifts, shift distribution balance, and nurse satisfaction, respectively, were 0.002, 0.01, and 0.001. These results indicate a statistically significant difference between the proposed algorithm's schedules and the traditional method. Additionally, night shifts were reduced by 38% compared to the supervisor's schedules.

The findings demonstrate that the proposed model can serve as an effective tool for optimizing nurse scheduling management by incorporating real-world hospital constraints and nurse preferences. Compared to traditional methods, it significantly improves shift fairness and enhances nurse satisfaction.