ESTIMATION OF ANNUAL AVERAGE DAILY TRAFFIC
USING NEURAL NETWORKS

S. Sharma¹, P. Lingras², and F. Xu¹

¹Faculty of Engineering
University of Regina
Regina, SASK. S4S 0A2

²Department of Computer Science
Algoma University College
Sault Ste. Marie, Ontario, P6A 2G4

ABSTRACT

Described in this paper is a comparison of neural network approach and the traditional factor approach of estimating annual average daily traffic from short period traffic counts. Minnesota’s highway traffic records are used for the purpose of this study. The results of this study indicate that 95th percentile estimation errors are highest in the case of factor model when applied to unclassified road sites. The neural network model using unclassified data provides much better estimation results. Road classification has little or no impact on the performance of neural networks. A factor model in which road sites are appropriately classified into various groups and the sample sites are correctly assigned to a known road group can provide better results than the neural network models. However, considering that a 100% correct assignment of all sample sites is extremely difficult, if not impossible, it is concluded that in actual practice, a neural network approach could easily outperform the traditional factor approach.

KEYWORDS: Average annual daily traffic, highway design, highway planning, traffic counts, traffic engineering, traffic volume