Analysis Of Driving Behavior And Traffic Congestion Based On Data Mining

With the rapid growth of the social economy and the sharp increase of car ownership,the problems of urban traffic safety and traffic congestion have become more and more acute.The intelligent transportation system as an information technology to solve a series of problems including traffic safety and traffic congestion emerged.Driving behavior is closely related to the traffic safety,traffic accidents caused by risky driving behavior are one important reason of traffic congestion.Therefore the accurate evaluation and the regulation of driving behavior are important measures to ensure the traffic safety;Effectively identifying the road traffic state and predicting the traffic status of the next time could provide an effective and scientific way for traffic managers to grasp the overall operation of the road,so the traffic managers can take measures for traffic evacuation and induction in a timely manner;On the other hand,traffic participants can choose the best way and time to travel,so as to reduce travel time and avoid traffic congestion sections.To sum up,driving behavior and traffic congestion analysis have improtant research values and significance in the application and development of ITS(Intelligent Transportation System).With the development of the technology of the internet of vehicles,the dimensions of information collected by the intelligent transportation system are more.In this paper,data mining is used to analyze the data collected by the intelligent transportation system.This paper is to provide the technical support and scientific basis for ITS.The main research achievements are as follows.(1)This paper puts forward an evaluation algorithm of driving behavior based on the combination of clustering and BP neural network.Firstly,an improved K-means clustering algorithm is used to make initial clusters of driving behavior;Secondly in accordance with the results of clusters,BP neural network classifier is trained;At last,the trained BP neural network is used to classify the driving behavior.The research result shows that the algorithm can eliminate subjective factors and make accurate,objective and efficient driving behavior evaluation,which provides scientific basis for traffic management department to monitor risky drivers and a new method of driving behavior evaluation for UBI(Usage Based Insurance).(2)In view of the fuzzy characteristics of traffic state,this paper proposes a real-time taffic state identification algorithm based on an improved fuzzy c-means clustering in which an adaptive fuzzy clustering model is introduced considering the different importance of each data point,and the simulated annealing algorithm(SA)is integrated with the particle swarm optimization(PSO)for improving the global search abilty of PSO to optimize the parameters of the adaptive fuzzy clustering model.Finally the speed,flow and occupancy are taken as feature attribute,a case study is conducted by the proposed real-time taffic state identification algorithm,and the results show the effectiveness and applicability.(3)In terms of the problem of the short-term traffic flow prediction,the nonlinear and delay properties of traffic flow are studied,and the method of determining the delay time of traffic flow system is given.Based on the grey system theory,aiming at the flaws of the grey prediction model,the grey prediction model is optimized and improved.Concerning the delay properties of traffic flow,this paper puts forward a grey delay and nonlinear traffic flow dynamic model based on the improved grey prediction model.Finally,the practicality and validity of the proposed moel are verified,research results show the proposed model can reflect the true condition of traffic flow system and improve the predictiong accuracy of traffic flow.