Traffic Forecasts. Inference Algorithm Based On Gibbs Sampling

In the 1950s, along with the automobile industry, traffic jams was starting to appear, and modern transportation science arose at the historic moment. For over 60 years, the man has understanding over the transportation operation law more clearly. People realize that this system is a complicated system, which has the obvious characteristics of randomness, dynamic and complexity. Based on Bayesian Networks reasoning knowledge, Markov blanket theory, and approximate reasoning theory, the thesis presents the traffic prediction technology which is based on Gibbs sampling reasoning.The thesis's innovation points are the thought of adding the relation between the travel time and the traffic congestion density (the equation about the vehicle flux, speed and driveway share) in the use of Gibbs sampling reasoning algorithm and the thought of the dynamic calculation of conditional probability, which may provide some reference for the later research.Firstly, the thesis briefly introduces the knowledge content of Bayesian network, also emphatically studies Markov blanket theory, approximate reasoning algorithm, and analyzes the fact that the traffic prediction can use Gibbs sampling reasoning algorithm. Then the thesis expounds Gibbs sampling algorithms thoughts, which lays the foundation for its improving application.Secondly, we process the primitive traffic data which is obtained from traffic bureau simply, and then based on the forefathers'achievements, we reason the relation among the traffic parameters in order to get the traffic congestion density, which plays a very important role in selecting the time difference between the adjacent nodes. Also this thesis realizes the calculation of the dynamic conditional probability, and forming the improvement in the application of Gibbs sampling reasoning algorithm, which lays the foundation for the traffic prediction.At last, studied on the nearby one two one street's traffic data, this thesis elaborates the original Gibbs sampling reasoning algorithm and the improved Gibbs sampling reasoning algorithm, and also it verifies the improved application of the Gibbs sampling reasoning algorithm is practical.