Different types of artificial neural network models were explored by researchers in the past. However, it is not clear from the past literature as to which of these models is best suitable for predicting the travel time of a link. Therefore, the objective of this study is to research and select the best neural network model structure to predict travel time on selected links. It is achieved by developing and comparing two-layer feedforward neural network model (neural network fitting), nonlinear autoregressive with external inputs (NARX) model, and nonlinear autoregressive model (NAR). Two links on I-85 freeway were selected for this study. The historical travel time data for the year 2014 and 2015 were collected from a private data source. The travel time was aggregated at 15-minute intervals. The developed models were tested by considering the travel time data for the year 2016. The results obtained indicate that NARX model outperformed the other two models, while NAR model performed better than the traditional neural network model for the selected links and data used in this research.
Link-level Travel Time Prediction Using Artificial Neural Network Models
2018-11-01
690472 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Travel time prediction with LSTM neural network
IEEE | 2016
|Comparing whole-link travel time models
Online Contents | 2003
|Comparing whole-link travel time models
Elsevier | 2002
|Link Travel Time Prediction Based on the Integration of Rough Set and BP Neural Network
British Library Conference Proceedings | 2009
|