Extending geographically-weighted regression from points to flows: a rail-based case study
Extending geographically-weighted regression from points to flows: a rail-based case study
At present in the UK an elasticity-based approach is used to forecast changes in rail passenger demand resulting from changes in both the rail service offer and in external conditions, with uplift factors calculated based on the proportional change in the level of explanatory variables over time. Changes in these explanatory variables may have differing effects on rail demand in different areas. This is currently controlled for via a limited segmentation of the market with different elasticities estimated for each segment, which inevitably limits the complexity of the variations which can be captured. This paper describes the use of Geographically-Weighted Regression (GWR) to enhance the modelling of such spatial variation. Firstly, conventional cross-sectional demand models were calibrated covering major rail flows across Great Britain. These models were then recalibrated using GWR to allow assessment of spatial variations in rail demand elasticities. Previous applications of GWR have almost exclusively focused on spatial data which have a single point location. This is not the case for rail flows, and the paper compares the results given by several different methods for defining point locations for flows. It also assesses different methods for approximating GWR results to simplify their application in real-life forecasting situations. The results show that the use of GWR can give a significant improvement in the fit of flow-based rail demand models, and that it is possible to spatially segment the UK passenger rail market based on the results from these models. In order to integrate such segmentations with the standard UK rail demand forecasting methodology it would however be necessary to extend the GWR methodology further to allow the calibration of GWR models on panel data.
723-733
Blainey, Simon P.
ee6198e5-1f89-4f9b-be8e-52cc10e8b3bb
Preston, John M.
ef81c42e-c896-4768-92d1-052662037f0b
2013
Blainey, Simon P.
ee6198e5-1f89-4f9b-be8e-52cc10e8b3bb
Preston, John M.
ef81c42e-c896-4768-92d1-052662037f0b
Blainey, Simon P. and Preston, John M.
(2013)
Extending geographically-weighted regression from points to flows: a rail-based case study.
Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit, 227 (6), .
(doi:10.1177/0954409713496987).
Abstract
At present in the UK an elasticity-based approach is used to forecast changes in rail passenger demand resulting from changes in both the rail service offer and in external conditions, with uplift factors calculated based on the proportional change in the level of explanatory variables over time. Changes in these explanatory variables may have differing effects on rail demand in different areas. This is currently controlled for via a limited segmentation of the market with different elasticities estimated for each segment, which inevitably limits the complexity of the variations which can be captured. This paper describes the use of Geographically-Weighted Regression (GWR) to enhance the modelling of such spatial variation. Firstly, conventional cross-sectional demand models were calibrated covering major rail flows across Great Britain. These models were then recalibrated using GWR to allow assessment of spatial variations in rail demand elasticities. Previous applications of GWR have almost exclusively focused on spatial data which have a single point location. This is not the case for rail flows, and the paper compares the results given by several different methods for defining point locations for flows. It also assesses different methods for approximating GWR results to simplify their application in real-life forecasting situations. The results show that the use of GWR can give a significant improvement in the fit of flow-based rail demand models, and that it is possible to spatially segment the UK passenger rail market based on the results from these models. In order to integrate such segmentations with the standard UK rail demand forecasting methodology it would however be necessary to extend the GWR methodology further to allow the calibration of GWR models on panel data.
This record has no associated files available for download.
More information
Published date: 2013
Organisations:
Transportation Group
Identifiers
Local EPrints ID: 353651
URI: http://eprints.soton.ac.uk/id/eprint/353651
ISSN: 0954-4097
PURE UUID: 770b3700-0ca1-4912-9bf7-a888a931bb10
Catalogue record
Date deposited: 12 Jun 2013 15:23
Last modified: 15 Mar 2024 03:32
Export record
Altmetrics
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics
