| At present, the criteria of highway traffic state are mainly focus on threshold comparison or absolute standards, however; the criterias ignore the impact of road environment, weather conditions and other objective factors under different temporal-spatial background. Thus, the highway basic sections are selected as targets and then the fuzzy clustering method which based on genetic algorithm is proposed. The method is utilized in determining the relative standard of highway traffic basic sections state identification. Therefore a decision model based on kernel extreme learning machine which has important resarch significance and application value in achieving the control and management of freeway traffic flow is built.The characterization parameters are selected by analyzing the relationship between operating and crowded characteristics of traffic flow and traffic flow parameters. Meanwhile, the relative criteria method of highway traffic basic sections state identification for solving the absolute standard defect of traditional traffic state discrimination is proposed. In addition, the clustering analysis of the historical traffic flow data is processed based on fuzzy C-means clustering algorithm. Then the acquired traffic state cluster centers are treated as relative standards for partition and the traffic state of the measured data is determined according to the Euclidean distance afterward. Considering the unstable of the algorithm and easily falling into local optimum in fuzzcy mean algorithm caused by the randomness of the selection of the initial cluster centers, the research improve the stability of fuzz mean algorithm and enhance the reliability of traffic state clustering through the genetic algorithm to optimize the cluster center selection. The real-time traffic status discrimination decision model based on nuclear extreme learning machine is proposed since the calculating time complexity of the Euclidean distance decision-making model is greater.Finally, the PeMS traffic characterization parameters measured operating data is used in testing the method of highways real-time traffic state identification. As a result, the Fuzzy C-means clustering algorithm based on Genetic Algorithm has good stability, fast convergence. On the basis of the guarantee classification accuracy, The Real-time traffic state decision model based on kernel extreme learning machine Greatly save the time cost, has good real-time performance. |
