Analysis And Prediction Of Traffic State Based On Spatiotemporal Trajectory Data

Traffic pressure has became a problem which every city has to face when developed into a certain extent.To mitigate the traffic pressure requires scientific and effective traffic management measures.It is the key content of the intelligent traffic management system that how to obtain the information and data in the traffic network to monitor the traffic state of urban road network in real time,and combine the massive real-time traffic information and the huge historical database of the road network,select the appropriate traffic parameters,establish effective mathematical methods,analyze the changes of traffic conditions timely and accurately,and then predict the traffic state in the future.These will help the traffic information services and to control and induce the traffic.This thesis used taxi GPS data which has high travel rates and coverage of road network as a high quality spatio-temporal trajectory data of floating car to reflect the traffic conditions of urban road network in real time,and used big data technology based on Hadoop plateform to manage and process the huge traffic state information to solve the massive data processing problems in the traditional way.It has the advantages of high efficiency,high accuracy and high timeliness,and this is important for analyzing and predicting traffic state in time.At the same time,it is not appropriate to describe and predict the traffic conditions only by the traffic volume due to the complex situation caused by the differences between different roads.In order to analyze and predict the traffic state on the road more effectively,accurately and timely,this thesis adopts road speed as the traffic state parameter.It is more reasonable and effective.In addition,this thesis analyzed the time series of road speed after road network division.The feature of quartile was used to optimize the algorithm thus improving the rationality and accuracy of the prediction method,and verifed its effectiveness through real historical data.The results showed that this method could both reflect the trend of road speed and weaken the influence of extreme values or outliers,and it also showed a reasonable changing process of traffic state.Besides,its calculation is simple.This point can effectively save computing resources in processing big data.The curve fitting of calculated results proved the reliability of quartile method in handling the road speed.At the same time,this thesis used a mathematical method based on weighted mean and correction values of historical data,to predict the future road speed of traffic.The results showed that the prediction method could effectively predict the trend of traffic state,and the predicted value of road speed was very close to the real value.The results of the prediction could contribute to traffic guidance services and traffic management decisions.