The Detection Of Traffic Congestion Based On Data Mining In Internet Of Vehicles

With the development of the economy of our country, the number of automobiles is growing so rapidly that it results in a series of problems about urban transportation and social issues where the traffic congestion seems particularly outstanding. How to detect traffic congestion effectively is a key technical challenge in Internet of Vehicles and Intelligent Traffic System. On the preprocessing the basic traffic information, the algorithm of detecting traffic congestion based on GPS data has been proposed in the thesis which mainly consists of three points as follows:Firstly, a method of transforming the traffic information to a pseudonormal distribution using a modified power function has been studied in this thesis in order to remove the abnormal data collected from the traffic information and enhance the accuracy of acquisition of traffic information. At the beginning, the traffic information is first checked for normality with a QQ- plot. If it does not satisfy normality, the data should be transformed to a pseudonormal distribution using a modified power function and then cleaned based on 3? rule. The simulation results show that the sample data processed by the proposed method mostly approximate the fitting curve.Secondly, A speed-based method of repairing data in the spatial-temporal correlation model has been proposed with regard to the continuous abnormal data in order to enhance the accuracy of repairing data. It aims to repair the missing data by combining the a secondary exponential smoothing method with the correlation of neighboring segments, The simulation results show that the proposed method is of greater quality for repairing data.Finally, A framework for detecting traffic congestion has been improved in this thesis. Firstly, the off-road clusters has been removed by the identification of on-road clusters algorithm. Then the theory foundation has been provided for making traffic decision through segregating the driving behavior models on the roads by segregation of distinct transportation modes algorithm. The simulation shows that this framework can help people avoid road congestion while traveling, minimize the travel cost and finally improve the operating efficiency of urban road network.