Research on a Combined Neural Networks Prediction Model for Urban Traffic Volume

Abstract

Urban traffic and transportation problems have become the main problem in the way of urban development. In order to resolve prediction problem of traffic volume, firstly, time series of traffic volume are reconstructed in the phase space in this paper, and correlative information in the traffic volume are extracted richly, then two-stage prediction system for traffic volume is applied: the first stage contains two parallel improved Elman neural networks, which are trained by standard back propagation algorithm, the second stage mixes prediction results of the first stage, which is trained by Karmarkarpsilas linear programming. Real example shows that predicted result of this method is famous, and this method has biggish applied potentials in the region of traffic control.