Study On Short - Term Traffic Flow Forecasting Method

With the development of intelligent transportation technology, vehicle inducing system as an important part of the intelligent transportation system has become an effective way to divert urban road traffic to the traffic management department. Short time traffic flow forecasting is a key technology of vehicle inducing system, besides, it is also a very important theoretical foundation in the field of intelligent transportation. Through the analysis and forecasting of road traffic flow, it can provide the optimal driving route, optimize the traffic management scheme as well as balance the traffic flow.Firstly, we analyze the background of short time traffic flow forecasting; research status home and abroad as well as the significance of topic research. On the basis of studying the principle of support vector machines, because support vector machine (SVM) has advantage in being better able to solve practical problems including high dimensionality, nonlinear, local minima points and so on, so we put forward using support vector regression (SVR) method to establish short time traffic flow forecasting model.Secondly, based on the study of SVM, use the square of training error instead of slack variables; improve inequality constraints into equality constraints. Then we put forward short time traffic flow forecasting model based on least squares support vector regression (LS-SVR). Because of avoiding solving quadratic programming problems, so it improves the speed of model trainer. Because the precision of the predictive results is influenced by the model parameters, in order to further improve the predictive precision of the model, we put forward to use fish swarm algorithm (AFSA) to optimize the parameters of LS-SVR, so we get predictive model which is based on AFSA-LS-SVR.Finally, we take the collected related traffic data as research subject which comes from Yangzhou Shuangqiao hillock September 4,2014 to September 6,2014. We respectively use SVR, LS-SVR, AFSA-LS-SVR predictive models to predict its short time traffic flow, and then compare and analyze the results obtained. The simulation results show that the predictive results of LS-SVR model is better than SVR model; and the predictive error of AFSA-LS-SVR model is smaller which shows that use AFSA to optimize the model parameters of LS-SVM model will help improve the prediction accuracy and AFSA-LS-SVR method has some advantages in aspects of forecasting. So the predictive method which is based on fish optimization algorithm least squares support vector regression has certain research value and social significance.