PREDICTION OF RECREATIONAL TRAFFIC ON SUNDAY AFTERNOONS
USING NEURAL NETWORK TECHNIQUES

P. Lingras¹, S. Sharma² and I. Kalyar³

¹Department of Computer Science
Algoma University College
Sault Ste. Marie, Ontario, P6A 2G4

^{2, 3} Faculty of Engineering
University of Regina
Regina, S.K., S4S 0A2

ABSTRACT

Prediction of traffic volume is important in collecting and transmitting information on traffic conditions and transmit schedules for travelers before and during their trips. Short-term predictions of traffic volume in the immediate future has significant need for an intelligent transportation system. This paper reports results of two methods to predict eastbound hourly traffic volume on Trans-Canada highway on Sunday afternoons in the province of Alberta. Different models were developed using regression analysis and artificial neural network methods. The two methods performed equally good when the prediction was done for all Sundays throughout the year. A regression model using data for summer months only could not be developed for predicting Sunday traffic during summer months due to insufficient data. No such limitation was encountered in the case of the neural network approach. The neural network models developed from summer data for predicting Sunday traffic during June, July, and August resulted in much lower prediction errors than the errors resulting from the models using the entire year data.

KEYWORDS: Neural networks, traffic congestion, traffic volume, recreational traffic, rural traffic, intelligent transportation system