We define metrics to quantify the level of overall delay and propose an agent-based data-driven model with four factors, including aircraft rotation, flight connectivity, scheduling process, and disturbance, to build a simulator for reproducing the delay propagation in aviation networks. We then measure the impact on the propagation by the delay at each airport and analyze the relevance to its temporal characteristics. When delay occurs, airline schedule planning may become infeasible, and rescheduling of flights is usually required to maintain the function of the system, so we then develop an improved genetic algorithm (GA) to reschedule flights and to relax the root delay. Results indicate that priority-based strategy rather than First-Come-First-Serve can achieve minimum overall delay when congestion occurs, and aircraft rotation is the most important internal factor contributing to delay propagation. Furthermore, the reschedule generated by the improved GA can decrease delay propagation more significantly compared to the agent-based model.