Short-term Traffic Flow Forecast By Improved SA-SVR Based On Spark

The rapid development of science and technology for people’s lives has brought convenience,but also brought some negative effects.Traffic accidents,road congestion,and global warming caused by vehicle exhaust emissions.These transport problems,as negative attachments to economic development,are one of the many problems that need to be addressed.Since the advent of the traffic problem,the study of the solution to the traffic problem has never stopped,and with the advent of the intelligent age,the concept of the intelligent transportation system has been raised.Intelligent traffic system as the current solution to traffic problems of choice,and short-term traffic flow as part of the intelligent transportation system is the researcher attention.The traffic flow is not static,it is a non-smooth non-linear system of interference with the external environment,and has a massive flow of traffic data.As time goes,the data are also growing.How to deal with these massive data and achieve the accuracy and real-time requirements of traffic flow forecast has become the main research direction in recent years.This paper studied how to improve traffic flow forecasting accuracy and real-time problem departure,the main research include:(1)Support vector regression(SVR)is proposed to deal with the small sample nonlinearity.Based on the study of the characteristics of traffic flow and traffic flow,this paper studies some short-term traffic flow forecasting models.And at the precondition of experiment.It has verified the feasibility and practicability of SVR as short-term traffic flow forecast.(2)Improved the simulated annealing algorithm(SA)for large-scale combinatorial optimization,then applied to support vector regression machine for parameter optimization.Based on the support vector regression machine,It has found that the parameters of the support vector machine have a crucial effect on the prediction results of the whole model.In order to achieve the short-term traffic flow forecasting model based on the optimal parameters,this paper studies and compared other traditional parameter optimization algorithms,establishes and improves the simulated annealing algorithm which is suitable for large-scale combinatorial optimization.Based on the improved simulated annealing algorithm,And the prediction model is established based on the optimal parameters,which solves the problem of prediction accuracy in short-term traffic flow forecasting.(3)The SA-SVR prediction model under Spark platform is established.With the increase of the amount of traffic flow data,the prediction model in the stand-alone mode can not meet the requirements of short-term traffic flow forecasting for real-time prediction due to the limitation of physical factors in the process of handling massive traffic flow data.In order to solve the problem of forecasting time,this paper studies the Spark technique with distributed parallel processing ability in the background of large data to do large-scale parallel algorithm training for support vector regression machine,and combines the support vector regression machine to deal with nonlinear The characteristics of the small sample event and the short time of the parallel processing time of Spark,the SA-SVR prediction model under the Spark platform is established.Experiments show that this model shortens the prediction time under the premise of ensuring the prediction accuracy,and satisfies the requirement of short-term traffic flow prediction for accuracy and real-time.In this paper,three sets of contrast experiments were carried out based on the prediction model,which are RBF neural network and support vector regression model,grid algorithm and simulated annealing algorithm and improved optimization model of simulated annealing algorithm.Comparison of Predictive Model Experiments with Spark Parallel Mode in Standalone Mode.The three sets of contrast experiments proved that the model based on the improved simulated annealing algorithm for parameter optimization of the support vector regression machine was more competitive in the Spark environment than the traditional algorithm and the prediction in the stand-alone mode.The model under the Spark platform not only solves the problem of the accuracy of short-term traffic flow forecasting,but also solves the problem of forecasting time in short-term traffic flow forecasting,and improves the traffic flow in the short-term traffic flow forecast’s ability and forecast accuracy and real-time.The main innovation of this paper is to combine the sparse features of the support vector regression machine with the parallel processing capability of the distributed cluster Spark system,and carry out the large-scale SVR training parallel algorithm under Spark,thus establishing the SA-SVR under the Spark platform,and the model solves the problem of accuracy and real-time of short-term traffic flow prediction.