Dynamic Scheduling of University Cultural Courses and Activities Through a Digital Twin-Based GNN and Reinforcement Learning Model

Cultural courses and events are becoming harder to administer at universities owing to overlapping schedules, limited venue and faculty availability, changing student preferences, and last-minute adjustments. To address these challenges, this work introduces a digital-twin-enabled dynamic scheduling framework that integrates Graph Neural Networks (GNNs) and Reinforcement Learning (RL). The digital twin mirrors real-time university cultural operations—events, resources, constraints, and participant behaviors—allowing the model to simulate, evaluate, and optimize scheduling decisions before deployment. The GNN–RL pipeline learns structural connections between events, resources, and participants, enabling repeated optimization that balances resource allocation, student happiness, and cultural programme diversity. Simulations within the digital twin demonstrate the system’s ability to respond to dynamic event updates, resolve conflicts, and allocate scarce resources more effectively than baseline heuristic and rule-based schedulers. Quantitative evaluations show significant improvements in utility gain (26.50%), student satisfaction (87.20%), resource utilization (91.50%), and scheduling efficiency (96.10%). The scheduling predictions were statistically validated with an MAE of 0.42, MSE of 0.18, and R² of 0.92, confirming their reliability. Overall, the findings indicate that the proposed digital-twin-driven approach is feasible, scalable, and data-driven for addressing the complex and dynamic scheduling needs of university cultural management.