Authors:
Nilgun Ferhatosmanoglu
1
and
Betul Macit
2
Affiliations:
1
University of Turkish Aeronautical Association, Turkey
;
2
Gazi University, Turkey
Keyword(s):
Forecasting, Airport Networks, TBATS, Regression with ARIMA Errors, Airline Passenger Volumes, Neighbour Effects in Modelling.
Related
Ontology
Subjects/Areas/Topics:
Applications
;
Data Mining and Business Analytics
;
Forecasting
;
Industrial Engineering
;
Methodologies and Technologies
;
Operational Research
;
OR in Transportation
;
Pattern Recognition
;
Software Engineering
Abstract:
Forecasting airline passenger volumes can be helpful for flight and airport capacity planning. While there are many parameters affecting the passenger volume, to our knowledge no work has directly studied the effect of neighbour airports in modelling of passenger volumes. We develop an integrated model for forecasting the number of passengers arriving/departing an airport, considering the airport’s interactions with its neighbour airports. In particular, we analyse the time series of the flights arriving to and departing from two largest airports in Turkey, namely Ankara Esenboga and Istanbul Ataturk Airports, and explore the interactions between these airports by using them as regressors for each other. We also apply independent models based on TBATS which was previously proposed in the literature to handle multiple seasonalities. In our experiments, TBATS performs better than ARIMA for independent modelling, and TBATS with multiple seasonal periods outperforms TBATS with single sea
sonality in majority of the cases. In several cases, the forecasting accuracy increases when the neighbour airports’ traffic data is used in modeling the passenger volumes.
(More)