Surgery Scheduling Optimization using an Adaptive Genetic Algorithm with Q-Learning Guided Tournament Selection

Scheduling surgeries in operating rooms is a critical task that directly influences patient outcomes, staff workload, and a hospital's ability to respond to emergencies. This paper proposes an adaptive genetic algorithm with a novel Q-learning-guided tournament selection method to adaptively select tournament sizes, enhancing the balance between exploration and exploitation during the optimization process. A domain-specific encoding scheme for patient-operating room-day assignments was employed to effectively represent and manage scheduling constraints. To evaluate the proposed method, a discrete-event simulation environment was developed to generate datasets based on benchmark scheduling patterns. Numerical experiments showed that the proposed approach consistently converged faster than a standard genetic algorithm and achieved an average reduction of 10.05% in total scheduling penalties, while maintaining zero delay for emergency cases. Furthermore, it outperformed other state-of-the-art genetic algorithm variants in identifying optimal solutions. Thanks to its adaptability and efficiency, the proposed method improved the scalability of surgical scheduling systems, facilitating real-world implementation and reducing operational costs.