Research On Short-term Traffic Flow Prediction Method Based On Least Square Support Vector Machine

In recent years,China has entered a stage of rapid development in both social economy and urban scale.The supply and demand between the increasing number of motor vehicles and limited road resources are difficult to balance,and the contradictions are becoming more acute.The following traffic problems have caused tremendous pressure on urban development.Traffic accidents,traffic congestion and environmental pollution have seriously affected people's daily life.Intelligent Traffic Systems(ITSs)has emerged precisely to alleviate these traffic problems,among which traffic control and traffic guidance technology are two relatively effective means.Timely and accurate traffic flow is the key to traffic control and guidance.However,traditional traffic flow prediction techniques often require specific model structures and assumptions,which are not enough to cope with increasingly complex traffic flow data,and some intelligent prediction algorithms combined with machine learning and other technologies such are emerging.The least square support vector machine(LSSVM)has demonstrated powerful capabilities for time series and nonlinear regression prediction problems if it can select appropriate parameters.In this paper,a short-term traffic flow prediction model based on LSSVM is proposed to accurately predict the traffic flow of a selected section in a certain period of time.Based on the pre-processed traffic flow data,the prediction model is built to verify the prediction performance,and the problems encountered in the establishment and implementation of the short-term traffic flow prediction model are studied.The research work of this paper mainly includes the following aspects:First,describe and analyze the basic parameters of traffic flow data,identify and correct the wrong and missing data according to the characteristics of parameters,and ensures that the traffic flow data used for research contains as much effective information as possible to improve the prediction accuracy.Second,least squares support vector machine(LSSVM),the least squares version of the standard support vector machine,is used for short-term traffic flow prediction to improve the long-term training time of the standard support vector machine and the high computational cost.Thirdly,in view of LSSVM's sensitivity to model parameters,a hybrid optimization algorithm based on particle swarm optimization(PSO)and genetic algorithm(GA)is proposed to optimize the parameters of LSSVM.In this paper,the crossover and mutation factors of GA are introduced into PSO,which makes the hybrid optimization algorithm not only retains the advantages of fast convergence and easy implementation of PSO,but also has the ability of GA to search the global optimal solution.Experimental results show that the hybrid optimization algorithm proposed in this paper can find a set of better parameters,so that LSSVM has good stability and prediction accuracy in traffic flow prediction.