Back to results list
Please use this identifier to cite or link to this item:
|Title:||Short-term traffic forecasting using Hong Kong annual traffic census||Authors:||Tang, Yuen-fan||Keywords:||Hong Kong Polytechnic University -- Dissertations
Traffic estimation -- China -- Hong Kong -- Mathematical models
Traffic surveys -- China -- Hong Kong
|Issue Date:||2002||Publisher:||The Hong Kong Polytechnic University||Abstract:||In Hong Kong, the Annual Traffic Census (ATC) report is used to present the statistical results of traffic volume on the automatic traffic counter stations in the middle of each year. In the ATC report, the Annual Average Daily Traffic (AADT) is the most useful traffic information for engineers/planners. In practice, for most on-going transport studies, there is always a need to use the most up-to-date traffic data such as AADT for model development and calibration. However, the current-year AADT data are not always available. Obviously, there is a need to continuously update the AADT based on historical ATC data and available current-year partial traffic counts. The short-term traffic flow forecasting is to predict the traffic daily and hourly volumes at a given location in the next time period of near future (say next day or next month etc) using historical and real-time traffic data. This can be incorporated into the Transportation Information System (TIS) or Advanced Traveller Information Systems (ATIS) to provide travellers with accurate and timely information so as to allow them to make the intelligent decisions on travel choices. In August 1999, the Traffic and Transport Survey Division (TTSD) of the Transport Department (TD) commissioned the Review of the Annual Traffic Census project in which works were carried on the investigation of new approaches for data analysis and development of user-friendly computer programs for computation and presentation. In this research, the short-term traffic flow forecasting models are investigated and the user-friendly ATC computer programs are enhanced. This thesis presents four models for short-term prediction of the hourly and daily traffic flows by the day of week and by month as well as the AADT for the whole current year. These four models are developed and based on the Auto-Regressive Integrated Moving-Average (ARIMA), Neural Network (NN), Non-Parametric Regression (NPR) and Gaussian Maximum Likelihood (GML) methods. The historical traffic data and available current-year partial traffic data are the input data used for model development/calibration. The results (both hourly and daily flows) of the four models are compared with the observed data for validation. The daily flows estimated by the four models are used to calculate the AADT for the current year. From the comparison results, the GML and NPR models appear to be more promising and robust for extensive applications for TIS or ATIS. Therefore, the GML and NPR models are incorporated into the enhanced ATC computer programs to provide the off-line short-term traffic forecasting database for the whole Territory of Hong Kong. The model validation is carried out using the up-to-date ATC data. The prediction results of the estimated AADT show that the average errors of all ATC core stations are less than +-10% from the NPR and GML models. The developed NPR and GML models could be used to predict the hourly and daily traffic flows and estimate the AADT for transport model development and calibration. It was found from the validation results that the NPR model is suitable for prediction of traffic flows in Hong Kong ATC stations. The NPR model is likely to react to unexpected changes more effectively than the GML model.||Description:||1 v. (various pagings) : ill. ; 30 cm.
PolyU Library Call No.: [THS] LG51 .H577M CSE 2002 Tang
|URI:||http://hdl.handle.net/10397/3455||Rights:||All rights reserved.|
|Appears in Collections:||Thesis|
Show full item record
Files in This Item:
|b1667764x_link.htm||For PolyU Users||162 B||HTML||View/Open|
|b1667764x_ir.pdf||For All Users (Non-printable)||9.31 MB||Adobe PDF||View/Open|
Citations as of Oct 22, 2018
Citations as of Oct 22, 2018
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.