Abstract: Global airline industry has been plagued by flight delays. During the flight operation, for the plane executing flight string tasks, the busy state of airport has a direct effect on the turnaround time of plane, further more effects the on-time departure performance of the consecutive flight. When constructing the prediction Bayesian Network(BN) model of delay propagation, the influence of airport busy state is considered which is described by aircraft movement per hour and whose relative factors are discovered through Bayesian structure learning. The ten times ten-fold cross validation results show that compared with the propagation BN model without consideration of airport busy state, the BN model considering airport busy state shows more accuracy in predicting departure delay.

Key words: flight delay propagation, Bayesian Network, flight string, ten-fold cross validation