Study On Traffic Flow Prediction And State Estimation Based On Support Vector Machine

Traffic flow guidance is considered as the optimum way to improve traffic efficiency and mobility, with the purpose of providing best route information for travelers in the transportation network. Traffic flow prediction and traffic state estimation are two key problems in traffic guidance system, with accurate real-time traffic flow and traffic state information, travellers can make suitable route choice. Therefore, it is necessary to predict traffic flow and traffic state information in the network accurately. This thesis focuses on the two problems.First, considering traffic flow prediction in a single road, due to the experiential risk minimization principle in neural network and other methods, it is faulty in theory, hard to determine network structure, overfitting and underfitting, local minima. Aiming at these problems, this thesis studies traffic flow prediction with support vector machine based on statistical learning theory. Besides, combining support vector machine method and data mining technology, importing weather state fator and work day pattern, speed and accuracy of traffic flow prediction are improved.Then, to the traffic state estimation problem, since traditional traffic state estimation methods divid the traffic state by the principles provided by the state traffic government, it is not suitable to different roads and hard to satisfy the real application needs. Considering this, in this thesis, fuzzy cluster method is used to divide traffic states. This method can divide different roads with different states. After this, multiclass support vector machine is applied to classify the traffic state in future time. Multiclass support vector machine is an improvement to the two class support vector machine. It can divide problems with more than two targets and is suitable for the traffic state estimation problem.Last, comibing traffic flow prediction and traffic state estimaton method, a real time traffic state estiomation system model is advanced in this paper. The system uses real time collected traffic parameter information to predict traffic state information in future time and distributes the traffic state information. It realizes traffic guidance well and efficiently improves the traffic system's service quality.