Prediction of Large-scale Flight Delays and Analysis of Delay Propagation Mechanism

Due to external uncertainties impacting the air traffic system and the interconnection of flights in both spatial and temporal dimensions, flight delays propagate across multiple airports, leading to a multitude of adverse effects. Therefore, predicting and analysis of flight delay propagation in advance can enhance airlines’ ability to assess the quality of upcoming flight schedules. To achieve this, we adopt a dynamic network perspective to investigate the propagation of air traffic delays. Specifically, we construct a spatial-temporal network. To analyze the real impact of delay propagation on next-day flights, we propose a deep learning weight model based on large-scale flight data for accurate delay prediction. Based on the above, we construct the evaluation metric to quantitatively assess the magnitude of delay propagation. To validate the effectiveness of the proposed method, we conduct a study using domestic flight data provided by East China Air Traffic Control (ECATC). Experimental results illustrate the method's capability to elucidate the details of delay propagation in Chinese air traffic within the planned air flight schedule. The metric provides a multifaceted assessment of delay propagation, aiding decision-makers in adjusting the air flight schedule. A new idea for studying air delay propagation is provided by this work.