Research On Urban Congestion Road Segments Prediction Based On Congestion Transition Matrix

In recent years, the increasing of birth rate and vehicles’ occupancy, the rapid development of vehicles’ number, and the lagging of road resources construction lead to more and more serious traffic congestion. Urban congestion road segments prediction becomes a fundamental function of intelligent transportation system. Predicting urban congestion road segments accurately, not only makes people choose a better route, but also reduces exhaust emission, so as to save time, money, energy and protect the environment. Although the research on congestion road segments prediction exits for a long time, it is still not perfect in term of speed, accuracy and scalability, especially the appearance of emergency congestion makes the accuracy of prediction reduce greatly.In order to solve the above problem, this thesis proposes a new model of urban congestion road segments prediction based on congestion transition matrix. The obvious difference between the model and the existing models is that it combines the traffic flow information on adjacent road segments for congestion road segments prediction and gets good prediction results. The main contributions of the thesis include:First, the thesis proposes a specific algorithm to set up the model of urban congestion road segments prediction against the characteristics of temporal and spatial correlation between two congestion road segments. It divides GPS dataset into different parts according to cycles and time frames and mines the congestion road segments and traffic flow transition probabilities within each time frame.Second, the thesis proposes the concept of congestion transition matrix, which records congestion transition probabilities among all the congestion road segments having the time and space correlations. The congestion transition matrix is created for each time frame, preparing for prediction.Third, the thesis puts forward a method of dealing with sudden congestion road segments. The method takes advantage of traffic flow transition probabilities between sudden congestion road segments and their adjacent road segments to predict the congestion road segments in next time frame, improving the accuracy of prediction.Finally, on the basis of full probability theory, the thesis proposes the method of using the product of congestion vector and congestion transition matrix to predict congestion road segments in a time frame or multiple time frames in the future.Experimental results show that the method for urban congestion road segments prediction is effective and efficient.