Exploring of causal delay relationships in airport networks based on a deep separable attention network
Exploring the causal relationship of the delay propagation in airport networks and obtain-ing the causal delay effect is significant in the analysis of the mechanism of the delay propagation in airport networks.This study proposes countermeasures for airport network delays.In order to scien-tifically and accurately explore the causal relationship of the delay propagation of airport networks,this study proposes a deep learning framework based on an attention mechanism,using convolutional neural networks.This framework is divided into two main components:the causal relationship explo-ration of depthwise separable attention-based dilated causal convolutional networks and the causality verification based on the permutation importance method.The concept of the delay propagation caus-al delay index is proposed to characterize the time steps required for delay propagation between air-ports.In this study,one time step corresponds to 1 h.A causal delay effect network diagram of air-port network delay propagation is then constructed.To further verify the effectiveness of the pro-posed method,operation data from domestic outbound flights in China during the summer,autumn,winter,and spring seasons of 2019 to 2020 are analyzed.The experimental results show that there is a wide range of delay propagation causality in China's airport network and that the causal relation-ship in the summer and autumn seasons is more than that in the winter and spring seasons.Moreover,the importance of airports in delay propagation is not consistent with its scale,as most of the"caus-al"airports in the causal relationship of delay propagation are small and medium-sized airports in China.Hence,attention should be paid to the improvement of the delay management capacities of such airports.In addition,owing to the causal delay effect of delay propagation,the average delay in-dex of delay propagation between airports in China's summer and autumn(winter and spring)sea-sons is approximately 4.5(5.6)time steps,which indicates that the delay of the"causal"airports in the summer and autumn(winter and spring)seasons will be transmitted to the"influenced"airports in 4.5(5.6)time steps and that the"influenced"airports can adjust their airport operation manage-ment measures in time,according to the delay index,to deal with the delay caused by the"causal"airports to thereby prevent the occurrence of large-scale flight delays.In summary,the results of this study can provide decision-making support for airlines,airports,air control,and other departments to reduce delays and improve the safety and efficiency of civil aviation operations.
air transportationcausality explorationdeep learningairport networksdelay propaga-tion
万方数据
维普
