Abstract: To apply the dynamic data-driven approach to the real-time prediction of flight delay, a state-space model of flight delay with the focus on the modeling of the variable of random delay is established based on analysis of the flight delay propagation effect in delay event sequence of a multi-task single flight, in order to represent the relationships among delay states as well as those between delay state and the random delay. Then case study is carried out on historical flight delay data. Expectation maximization (EM) algorithm is applied to calculating the maximum likelihood estimation of the model parameters, and the model is validated by probability intervals test on different instances. Test results show that the proposed state-space model of flight delay has good generalization performance.