Traffic Flow In Normal And Abnormal Prediction Using Time-Space Fusion Model Based On Support Vector Machines

With the high-speed development of the national economy and the urbanization, urban transport development has been rapid progress. However, improvement of living standards of urban residents, the rapid growth of vehicle ownership, leading to between surge in road traffic and the limited road resources continue to intensify the contradiction causing a series of traffic problems, such as the more serious traffic congestion, traffic accidents and the environment pollution and so on. The development of domestic and international experience has shown over the years, any country or region can solve traffic congestion problems by large-scale road construction. Under current conditions, only the rational use and maximize exert the potential of urban road network, can be integrated and coordinated balance between vehicles and roads. The solution to this problem is to correctly evaluate the basis of the current work status of urban road network, as defined by quantitative analysis of the reliability of calculations to evaluate the level of traffic. As one of the basic indicators of traffic engineering, traffic volume predict can not be ignored.Traffic predict for the current study generally ignored the interdependence at the same time between different sections. From the perspective of the entire road network, congestion or failure of the upstream and downstream sections, the section associated with will be affected. Therefore, reliability analysis which considered the relationship between sections is necessary. This paper presents a support vector machine (SVM) fusion of concurrent two-dimensional space-time predict method of traffic flow in two parallel system model to reduce the time cost. Also considered the relevance between time and space, on two-dimensional space-time fusion, greatly improving the predict accuracy. It can be effectively evaluate the reliability of travel time to provide more accurate data to support.This paper analyzes the international issues of common urban transport firstly, and make a general description of the basic principles and requirements for the effective way of evaluation of transportation system currently--- network reliability analysis. As the reliability analysis of network traffic requires a lot of projections, which leads to the contents of this research - the state of normal and abnormal traffic flow predict. By researches the basic theory of SVM research and development, and selects the methods based on support vector machine regression model to predict the traffic flow under normal and abnormal in space-time two-dimensional fusion model, and compares with the results of multiple regression method under normal and abnormal state, you can visually see the two-dimensional fusion model based on support vector machine shows better performance.