In order to accurately predict the flight time of aircraft during the departure phase and improve air traffic management efficiency, the factors influencing the flight time of departing aircraft are explored by extracting characteristic waypoints and multiple stacked model algorithms are established to optimize the prediction of flight time for departing aircraft. The base learners in this model integrate neural networks, support vector machines, and linear methods to enhance prediction accuracy, while the meta-learner employs a Bayesian ridge regression model to optimize the final results. The results indicate that the stacked model outperforms single base learners in flight time prediction tasks, with an average relative improvement of 18.8%, demonstrating stronger generalization capabilities and accuracy.