Journal of Highway and Transportation Research and Development · 2005 · 20 citations · 0 references
EngineeringComputer ScienceForecastingMachine LearningTransportation EngineeringIntelligent Transportation SystemsSupport Vector MachinesTime Series PredictionTraffic EngineeringTraffic Flow AnalysisTraffic Flow TheoryMachine Learning For Traffic AnalysisMachine Learning-based ForecastingReal-time Traffic ForecastingSequential OptimizationTraffic ControlTraffic Flow ForecastingTraffic Flow PredictionTraffic Prediction
The accurate prediction of real-time traffic flow is essential in intelligent transportation system(ITS).The prediction results will have direct effect on traffic control and traffic guidance.A traffic flow prediction model using support vector machines(SVMs) based method is proposed.With the sequential minimal optimization(SMO) algorithm,the proposed model can predict traffic flow efficiently.The simulation results show that the SVMs can avoid over-fitting and have better generalization ability compared with BP or recurrent neural networks.The average prediction error is 3.13% if there is no noise in the inputs.When the measurement inputs are disturbed by noise accounting for 10% of the input,the designed SVMs can keep robust and have an average prediction error up to(4.25%.)