Flow based spatio-temporal graph network model for predicting airport network delays
Résumé
With the continued advancement of the air transport
industry, ensuring the safety and efficiency of air traffic operations
has become a major concern. Various efforts are being made to
achieve this objective, among which the prediction of delays in
air traffic is of significance. Current deep learning methodologies
for predicting network wide air traffic delays typically rely on
historical delay data as the primary input, often neglecting the
inclusion of air traffic flow data. In this paper, the FSTGMAN
model is developed to explore the efficacy of incorporating traffic
flow data as an additional input for predictive modeling, contrasting
its performance with a model that does not utilize flow data.
The findings reveal that the incorporation of flow data marginally
enhances the overall accuracy of the predictions. Furthermore, the
performance of our model is compared with that of baseline models
such as MLP, LSTM, Transformer, and Seq2Seq, demonstrating
notable advantages.
Origine | Fichiers produits par l'(les) auteur(s) |
---|