A methodology for flight departure delay prediction based on stacking technique

Predicting flight delays helps reduce financial losses and increase passenger satisfaction. In this paper, the stacking technique is used to predict departure flight delays. Firstly, a feature set is constructed including flight attributes, weather, cyclical data, crowdedness degree of the airport, and air route situation, based on the previous studies. Secondly, five models are used to predict flight delays. Furthermore, the stacking technique is employed to improve prediction performance. Finally, the data from the departure flight from Beijing Capital International Airport (PEK) is used to be a case to validate the delay prediction model. The results show that the stacking technique outperforms other algorithms in prediction performance.