Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.814647
Title: Developing advanced methods to predict air traffic network growth
Author: Busquets, Judit Guimera
ISNI:       0000 0004 9354 7020
Awarding Body: City, University of London
Current Institution: City, University of London
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
This dissertation describes a forecasting methodology that takes into account changes in the connectivity of an air transportation system and assesses the impact at other levels of the network, such as route demand and air traffic levels. To achieve this, the modelling framework looks at city pair demand generation, route demand assignment and air traffic estimation. While generating air traffic forecasts, the resulting model is also intended to highlight the most important factors driving air traffic network growth. This is achieved by considering a larger set of drivers than those considered in existing methodologies and research as well as exploring the use of alternative modelling techniques. Network evolution is incorporated in the method through an airport connectivity model which identifies how and when airport-pairs across the network change their connectivity status. The problem is split into two models: one identifying those airport-pairs that are added to the network; and another one identifying those airport-pairs that are removed from the network. The modelling approach explores the use of network theory metrics along with other input variables, such as passenger demand, to see whether existing models employing only network theory metrics could be improved. The impact of network evolution is assessed by the effect on air itinerary shares. Two itinerary choice models are developed using two different modelling approaches: multinomial logit and neural networks. While the multinomial logit formulation is the most common approach used to model itinerary shares, only few studies have looked at modelling itinerary shares at the network level. Neural networks have yet to be explored in this field. In this research, air itinerary choice models have been developed at the most aggregate level, using open-source booking data, for a large group of city-pairs within the US Air Transportation System. The output of the itinerary choice models, influenced by the consideration of network evolution, is then used to project air traffic levels and assess the impact of network structure changes in the number of operations in the US ATS. The results reflect the complexity behind network evolution, especially for cases when a mature system is considered (e.g. US ATS): comparisons between the case of a static network and the case when network evolution is considered indicate that the impact of network changes on overall system metrics is relatively minor in the US. However, they indicate that changes in fossil fuel prices may influence changes in the overall network characteristics, and consequently network evolution. The results also prove the feasibility of estimating a single itinerary choice model at the network level for an entire air transportation system. Although the multinomial logit model results have better accuracy, the potential of neural networks for this purpose is also demonstrated, the latter being more representative of the hub-and-spoke network strategy.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.814647  DOI: Not available
Keywords: TA Engineering (General). Civil engineering (General)
Share: