Munich Personal RePEc Archive

Airport and Access Mode Choice in Germany: A Generalized Neural Logit Model Approach

Gelhausen, Marc Christopher (2006): Airport and Access Mode Choice in Germany: A Generalized Neural Logit Model Approach. Published in: Proceedings of the 2006 European Transport Conference (2006): pp. 1-32.

WarningThere is a more recent version of this item available.
[img]
Preview
PDF
MPRA_paper_4236.pdf

Download (15MB) | Preview

Abstract

The purpose of the paper is to present a novel approach of a general airport and access mode choice model. Based on data of the German Air Traveller Survey 2003 with a sample size of about 210.000 passengers interviewed at 21 airports a three-stage nested logit model has been estimated in a first step. 7 different access modes to the airport are modelled, subdivided into four private and three public travel modes. The model includes 7 different market segments: Domestic, European and Intercontinental travel, each segment split up into private and business travel. The European private travel segment is further subdivided into short stay trips and holiday travel.

The aim is to develop a generally applicable airport and access mode choice model. Thereby it is possible to analyse future in terms of new airport constellations and new airport access modes. To achieve this, Kohonens Self-Organizing-Maps are used to identify different airport clusters and assign every airport to the appropriate cluster. Based on these airport clusters the aforementioned nested logit model has been estimated.

In a second step, neural networks are applied to the problem of airport and access mode choice. On the basis of neural networks a new kind of discrete choice model called "Generalized Neural Logit Model" has been developed. To optimize the network structure genetic algorithms have been applied. Such a model fits into the structure of a General Extreme Value model and satisfies the condition of utility maximization.

A second airport and access mode choice model based on the Generalized Neural Logit Model and the airport clusters has been estimated. Although the former approach showed for most market segments a good model fit, the new approach showed a significant increase in model fit especially for those market segments the model fits of which in the nested logit model were less satisfying.

Available Versions of this Item

UB_LMU-Logo
MPRA is a RePEc service hosted by
the Munich University Library in Germany.