Research On The Technology Of Finding And Predicting The Areas Of Traffic Congestion

From the traffic situation in the country or even in the world, we can see that with the increasing number of motor vehicles in various countries, the existing public transportation vehicles will become unable to meet the growing capacity of the various transportation systems. Nowadays traffic jams become a universal existence and has been a serious problem for the people and society which can not be ignored. The best way to solve this problem is to "defend", to prevent the problems of traffic congestion before it happening. In addition, we analyze the problems in the ITS system in this paper, which is a waste of data resources.Give that data resource waste problem for the traffic congestion and intelligent transport systems, this paper firstly used the data which collected by the intelligent transportation systems to identify traffic congestion temporal region, followed by the various regions predict traffic jams occur at a later time based on the results found in the area of traffic congestion probability situations. The main purpose of this paper is to find out the traffic congestion area, and predict the probability of congestion based on the result of congestion area’s finding. The data used in this paper as a data source is collected by ITS systems which is the feedback from the Beijing taxi in November,2012.Firstly, considered the data noise caused by the GPS location system and caused by the principle of the system, and the data sets must be cleaned first and time slice. Secondly, cluster the time slice data sets based on distance clustering algorithm(Kmeans, DBSCAN) based on intensive features, and compare the results of the clustering methods. The DBSCAN algorithm is selected to analyze the data sets of each time slice according to the comparison results. Match the results of clustering algorithm and segmented regions, and the region can be divided into inside areas of clustering data, the edge of clustering data cluster, outside of clustering data cluster. In this paper, the average speed of vehicles in each region will be calculated, the speed of the outside of clustering data cluster will be recorded as zero. And then this paper accorded to the regional congestion decision rules to find out each areas of traffic congestion of each time slices. The final results are stored in a text file by the form of matrix.Because of the complexity of traffic congestion, there is a certain randomness of traffic congestion. Then the traffic congestion can be seen as the current state which only depends on the previous state, so this paper build traffic congestion prediction model based on Markov chain model. And a part of the discovery of regional congestion results as a priori knowledge which is the sample set too, and the rest of the discovery of regional congestion results as the validation set.Finally, the K-means/DBSCAN clustering, based on the Markov chain forecasting model are tested and compared in this paper, and the accuracy of the prediction model is verified.