SHORT TERM LOCAL ROAD TRAFFIC FORECAST USING FEED-FORWARD AND RECURRENT ANNS

The subject of short term municipal traffic prediction has been considered. Artificial neural networks (ANNs) were implemented for prediction. Starting with the hypothesis that one needs two predictions supporting each other, in order to be confident in the obtained result we have implemented two different ANN structures. These structures were earlier developed for and successfully implemented to short term electricity load prediction at suburban area, a problem having inherent similarities with local municipal road traffic flow. The final prediction was obtained as an average of the two predictions. A new algorithm is proposed for estimation of the number of hidden neurons in each of the two ANN structures.