Abstract: As there exists a link between the scheduled flights, each scheduled flight delay can affect the downstream airport and its scheduled flight.Therefore methods are needed to investigate and visualize the propagation of delays for flightconnection at hub airports. The Bayesian network (BN) is an effective propagation analysis method. In this article, a flight delay propagation model is developed based on the BNs by analysis of a certain hub airport flight delays which will affect associated airports and connecting flights. By utilizing some actual airline data, the article reports training of the model through the expectation maximization (EM) algorithm, and gives the test results, which clearly demonstrate the value of BN for analyzing the system-level effects arising from micro-level causes.

Key words: Bayesian network, flight delay, delay propagation, flight connecting, associated airport