الفهرس 
INTRODUCTOINOnly 14 pages are availabe for public view
 |  | from | 216 |  |  |
| | | from | 216 | | |
Abstract
Traffic volume is a fundamental measure that is used to serve different types of transportation-related applications at planning and operational levels. Traffic data collection is usually costly and hence, researchers started to adopt approaches to compensate for the scarcity of data and limited resources. In the past few decades, many effective approaches have been developed and applied to estimate/predict traffic volumes. Time series analysis is one of these valuable approaches where future values of traffic volumes at a facility are estimated using the time-relationships with past values. This research investigates the application of Box-Jenkins time series models to predict daily and weekly traffic volumes for a group of national Egyptian highways and to help in the analysis of seasonal variation on desert roads.
Five years of traffic volume data were collected between 2008 and 2012 at 14 different count stations and were used in the analysis. Daily and weekly traffic volumes for the years 2008 till 2011 were used to fit time series models. Seasonal autoregressive integrated moving – average (SARIMA) models were shown to be the most suitable modeling structure for the analysis. The developed models were used to forecast traffic volumes for the year 2012. The forecasted traffic volumes were then compared to the actual traffic volumes and the estimation error was computed. In general, estimation errors ranged between 4.3% and 7.10% for daily traffic volumes, while the error ranged between 2% and 7% for weekly traffic volumes.
The results of this research show the potential of using time series models for predicting daily and weekly traffic volumes on the national highway network of Egypt. The results also show that there is no single model structure that will best fit the data of all highway types. Virtually every highway project needs a traffic forecast to help define the scope, the geometry of the project and sometimes even help determine if a project is needed. Models developed in this research would be useful for monitoring and updating transportation plans. A related application of time series models would be their ability to supply quick and accurate forecasts would be of assistance in optimal control operations where decisions are being made continuously and automatically from predicted variables like demand. This is typical in traffic signal control systems.