En Route Congestion Prediction Method for Air Route Network Based on Spatiotemporal Graph Convolution Network and Attention

En route traffic congestion can lead to serious flight delays, affecting the operational efficiency of airlines, increasing the complexity of air traffic control, and reducing passenger air travel comfort. Effective prediction of en route traffic congestion nodes and their changing trends is a research hot spot in the field of air traffic management. This article presents a novel framework for prediction of en route traffic congestion. Initially, an en route congestion indicators (CIs) system is established to describe the congestion status of air routes. Subsequently, the attention-based spatial–temporal graph convolutional network (ASTGCN) model is employed to forecast CIs of the air route network. This model integrates an attention mechanism to capture the spatiotemporal correlations within the air route network while employing graph convolution to extract the network’s spatiotemporal features. Finally, a congestion level identification model based on the fuzzy <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">c</i>-means light gradient boosting machine algorithm is devised. Experimental validations are conducted on a selected air route network in the central and southern regions of China. The findings reveal that compared to baseline models, the ASTGCN exhibits the most robust and satisfactory performance in predicting en route CIs, with superior performance in terms of the mean absolute error, root-mean-square error, and mean absolute percentage error within a 6-h prediction window. Furthermore, utilizing the predicted CIs, the proposed model achieves congestion level identifications with a maximum error of one. The accuracy and precision of congestion level identification reach 84.76% and 83.47%, respectively, underscoring the practicality of the proposed en route congestion prediction approach.