2010, Vol.13, No.1, pp.53-63

The road traffic is considered as a non-autonomous dynamic phenomenon and modeled statistically by a non-parametric approach. Information for the modeling is extracted from recorded time series of traffic flow rate and the day-code specified by the calendar. An optimal model of the traffic flow rate is formed by the conditional average estimator. The condition is comprised of hour, day-code, data of past flow rate and weather variables. As an example the traffic flow rate at a representative point on a high-way in Slovenia is modeled. The model is further utilized to forecast the traffic flow rate. The forecasting performance is evaluated by the correlation coefficient r of the forecasted and original data. The mean value indicates surprisingly good modeling. The mean value depends on the combination of variables comprising the condition and could be optimized by changing their composition. The modeling is utilized in a graphic user interface by which parameters of forecasting can be selected and its result displayed. The mean traffic flow rate record over a week exhibits a characteristic structure that can be well represented by a superposition of normal distributions. Their parameters represent new information about the traffic phenomenon and related activity of population.
Key words: traffic flow forecasting, non autonomous chaotic time series, nonparametric regression

Full text:  Acrobat PDF  (706KB)  

Copyright © Nonlinear Phenomena in Complex Systems. Last updated: June 8, 2010