Neural Network Based Route Weight Classification and Prediction for Traffic Management System

Abstract

Traffic jam is a major problem in Dhaka City, so a traffic management support system, with less cost, flexible, easily maintainable and secured is in demand. For monitoring road traffic condition, Internet based real time bi-directional communication provides a lot of benefits. For making traffic system more realistic and reliable, dynamic route computation is a vital requirement. Therefore, for predicting road weights, an integrated approach with multiple data feeds and back propagation neural network with Levenberg Marquardt optimization is applied. The traffic system where NN based dynamic weights computation is used and much more suitable to find the optimal routes. Inclusion of BPNN with LM achieved more than 90% accuracy. NARX time delay neural network is used to predi����t differe����t feature’s ����eights a����d those are applied in this neural network to determine the road weights of different roads. NARX neural network performs better than weighted mean moving average to predict different feature’s ����eights.