Urban Traffic State Identifiction And Prediction Based On Genetic Alorithm Optimized Support Vector Machine

The traffic volume of existing urban road networks is difficult to match the rapid growth of car travel,which leads to frequent traffic congestion incidents in cities and seriously affect people’s normal life and work.The immature urban traffic management system is the root cause of traffic congestion.Furthermore,it is extremely urgent to build an efficient and intelligent urban traffic management system to solve urban traffic congestion problems.The identification and prediction of urban road traffic status is the basis and premise of establishing intelligent traffic management system.Therefore,in order to provide theoretical support for the construction of intelligent transportation system,this paper conducts in-depth research on urban road traffic status identification and prediction,.Based on the traffic flow basic parameters collected by the urban microwave detector device of Yuelu Avenue in Changsha City—flow,speed and occupancy rate,the K-means algorithm in the cluster analysis algorithm is used to classify and evaluate it,and construct the smooth flow and the stable flow and the crowded flow and the blocked flow are the basic data sets for urban traffic state identification of the evaluation indicators.On this basis,the genetic algorithm and support vector machine are combined to optimize the penalty coefficient and kernel function of support vector machine,and the urban traffic state recognition model based on GA-SVM is constructed.The model has better recognition accuracy than SVM model,the test set recognition accuracy rate reached 98.75%.Based on the short-term traffic flow theory,the traffic flow data of the local section of the Fourth Ring Road in Beijing is collected.Considering the short-term traffic flow prediction research,the spatial correlation is less studied.This paper applies the advantage of the strong global search ability of genetic algorithm to the parameter optimization of support vector regression machine,and solves the problem that the support vector regression machine is difficult to accurately select the parameters during prediction.A GA-SVR shortterm traffic flow prediction model based on spatiotemporal fusion is constructed.The traffic flow predictions in the early,middle and late periods are analyzed and compared with the traditional SVR prediction model.The prediction error based on the space-time GA-SVR model is relatively stable and the generalization ability is stronger.Through the quantitative analysis of the prediction results of the two models and the actual flow by the mean square error,the average absolute error and the average absolute percentage error,it is found that the GA-SVR model based on spatiotemporal fusion has higher prediction accuracy.