Modeling multi-system coupled risk propagation and AI-driven adaptive emergency resource allocation for large-scale sporting events

Large-scale sporting events are complex socio-technical systems in which local disturbances may propagate across crowd, transportation, facility, security, and medical subsystems. Traditional static risk assessment methods are limited in describing such cross-system cascading processes and their implications for emergency resource allocation. This study develops an exploratory simulation framework that combines weighted complex-network representation, directional subsystem coupling, resilience evaluation, and Deep Q-Network (DQN)-based adaptive resource scheduling. The event environment is represented as a synthetic, domain-informed 24-node network across five functional subsystems; therefore, the numerical results should be interpreted as mechanism-level simulation evidence rather than venue-specific forecasts. Under the baseline scenario, coupling effects increase the peak system-wide risk by approximately 21%, and the overall risk exceeds the warning threshold within 18 time steps if no intervention is implemented. Compared with the no-intervention baseline, the DQN-based adaptive scheduling strategy improves the composite resilience index by 39.5% and reduces the peak risk to approximately 48% of the baseline level. These findings suggest that cross-system coupling and adaptive resource allocation are important considerations for simulation-based safety management of large-scale sporting events, while future work should calibrate the framework using venue-specific operational data.