FuzzyNeural Network ModelApplied intheTraffic FlowPrediction

Thepaperproposes a fuzzyneural network themodelsofKalmanfilter, MovingAverage (MA)and model(FNNM) strategy forpredicting thetraffic flowofreal Autoregressive Integrated MovingAverage (ARIMA)(8). timetraffic control systems. Theproposed modeliscomposed Despite thefactthattheneural network iseffective for oftwomodular. Oneisafuzzy network (FN), whichisused short-term prediction oftraffic flow, itusually requires long forfuzzyclustering. Eachcluster represents onekindof training timeforthenetwork, especially whenthetraining specific traffic pattern. Theother isaneural network (NN), parameters, suchastraining rate, momentumandinitial whichisone-layer network andisusedforpartitioning the weights, arenotproperly chosen. Thisconfines its application relationship ofinput andoutput vector. AndtheFNmodule fortheonline training procedure fortraffic flowprediction. supervises thelearning oftheNN.Thatis, thefeatures ofthe Inthis paper, wepresent atraffic flowprediction model traffic samples areemployed toguide thetraining oftheNN. using thefuzzy neural modelwithon-line iterative algorithm Moreover, an on-line iterative predictive algorithm is topredict thedownstream traffic flowoftheintersections presented inthis paper topredict thetraffic flowaccording to according totheupstream traffic flowdata. Thereal sampled thesampled dataoftheupstream cross roads. Finally, thereal datafromcertain intersection isapplied toverify the sampled traffic flowdata isemployed tovalidate theproposedeffectiveness ofthepresented model. Experiment results show method. Results showthat theproposed traffic flowprediction thattheproposed methodcantrack thedynamics oftraffic strategy based onfuzzy neural network modelisfeasible and flow. Andthis method iseffective andpractical.