Passenger flow forecasting framework based on vision transformer and inpainting: Application to a public transport system
Résumé
Short-term forecasting is one of the most important challenges in intelligent transport systems (ITS). The demand information is crucial for transport operators in order to anticipate and optimize their service level and for travelers to have robust information. In addition, a good predictor can contribute to system resilience by predicting disrupted situations. In most forecasting models, the data collectors are based on regular time series and single stations of the urban area. This study considers trains as data collectors, i.e., sensors. Thus, we are not limited to single location sensors. Moreover, we have to deal with irregular time series.
Domaines
Autre
Origine : Fichiers produits par l'(les) auteur(s)