Application of generalised regression neural networks in trip distribution modelling
Trip distribution is the second step of the transport modelling process. Errors in this trip distribution step will propagate through the other stages of the transport modelling process and will affect the reliability of the model outputs. Therefore, finding a robust and efficient method for trip di...
| Main Authors: | , |
|---|---|
| Format: | Journal Article |
| Published: |
2014
|
| Online Access: | http://search.informit.com.au/documentSummary;dn=805587695517223;res=IELENG http://hdl.handle.net/20.500.11937/32785 |
| _version_ | 1848753759952306176 |
|---|---|
| author | Rasouli, M. Nikraz, Hamid |
| author_facet | Rasouli, M. Nikraz, Hamid |
| author_sort | Rasouli, M. |
| building | Curtin Institutional Repository |
| collection | Online Access |
| description | Trip distribution is the second step of the transport modelling process. Errors in this trip distribution step will propagate through the other stages of the transport modelling process and will affect the reliability of the model outputs. Therefore, finding a robust and efficient method for trip distribution has always been an objective of transport modellers. The problem of trip distribution is non-linear and complex. Neural networks (NNs) have been used effectively in different disciplines for solving nonlinear problems. Accordingly, in this paper, a new NN model has been researched to estimate the distribution of the journey to work trips. This research is unique in two aspects: firstly, the training of the model was based on a generalised regression neural network (GRNN) algorithm, while the majority of previous studies have used a backpropagation (BP) algorithm. The advantage of the GRNN model over other feed-forward or feed-back neural network techniques is the simplicity and practicality of the model. The second unique aspect is that the input data for the GRNN model was based on land use data for each pair of zones and the corresponding distance between them, while the previous NN models used trip productions, trip attractions and the distance between a pair of zones as inputs. As a case study, the model was applied to the journey to work trips in the City of Mandurah in Western Australia. The results of the GRNN model were compared with the wellknown doubly-constrained gravity model and the BP model. |
| first_indexed | 2025-11-14T08:29:38Z |
| format | Journal Article |
| id | curtin-20.500.11937-32785 |
| institution | Curtin University Malaysia |
| institution_category | Local University |
| last_indexed | 2025-11-14T08:29:38Z |
| publishDate | 2014 |
| recordtype | eprints |
| repository_type | Digital Repository |
| spelling | curtin-20.500.11937-327852017-01-30T13:33:04Z Application of generalised regression neural networks in trip distribution modelling Rasouli, M. Nikraz, Hamid Trip distribution is the second step of the transport modelling process. Errors in this trip distribution step will propagate through the other stages of the transport modelling process and will affect the reliability of the model outputs. Therefore, finding a robust and efficient method for trip distribution has always been an objective of transport modellers. The problem of trip distribution is non-linear and complex. Neural networks (NNs) have been used effectively in different disciplines for solving nonlinear problems. Accordingly, in this paper, a new NN model has been researched to estimate the distribution of the journey to work trips. This research is unique in two aspects: firstly, the training of the model was based on a generalised regression neural network (GRNN) algorithm, while the majority of previous studies have used a backpropagation (BP) algorithm. The advantage of the GRNN model over other feed-forward or feed-back neural network techniques is the simplicity and practicality of the model. The second unique aspect is that the input data for the GRNN model was based on land use data for each pair of zones and the corresponding distance between them, while the previous NN models used trip productions, trip attractions and the distance between a pair of zones as inputs. As a case study, the model was applied to the journey to work trips in the City of Mandurah in Western Australia. The results of the GRNN model were compared with the wellknown doubly-constrained gravity model and the BP model. 2014 Journal Article http://hdl.handle.net/20.500.11937/32785 http://search.informit.com.au/documentSummary;dn=805587695517223;res=IELENG fulltext |
| spellingShingle | Rasouli, M. Nikraz, Hamid Application of generalised regression neural networks in trip distribution modelling |
| title | Application of generalised regression neural networks in trip distribution modelling |
| title_full | Application of generalised regression neural networks in trip distribution modelling |
| title_fullStr | Application of generalised regression neural networks in trip distribution modelling |
| title_full_unstemmed | Application of generalised regression neural networks in trip distribution modelling |
| title_short | Application of generalised regression neural networks in trip distribution modelling |
| title_sort | application of generalised regression neural networks in trip distribution modelling |
| url | http://search.informit.com.au/documentSummary;dn=805587695517223;res=IELENG http://hdl.handle.net/20.500.11937/32785 |