Modeling and Estimating for Flight Delay Propagation in a Reduced Flight Chain Based on a Mixed Learning Method

Flight delay and delay propagation has been paid more and more attention by the Civil Aviation Administration of China (CAAC). Flight delay is the source of propagation, while delay propagated within a Flight Chain. Busy hub-airport plays an important role in a Flight Chain, and the Initial Delay often happens there. Through analyzing delay status of the busy hub-airports in a Flight Chain, the status of whole chain will be found out basically. Bayesian network (BN) is chosen as the tool to model and estimate flight delay in a busy hub-airport. We proposed two modeling methods with different algorithms, which are separately based on parameter learning and structure learning of BN. The models learned by K2 provide a successful topology for estimating the flight delay, with all the estimating correct rates are higher than 90%. Then we use a method mixed by the both structure learning and pure parameter learning to build a network model for a reduced Flight Chain, the modelpsilas structure is established based on the learned topology. The delay estimation by the model proves much better than the old model trained by pure parameter learning.