Truck's Risk Assessment And Influencing Factors Based On Vehicular Networking Data

In recent years,the incidence of truck accidents and the death rate caused by truck accidents in China have remained high,the safety problems in the transportation process need to be solved urgently.With the development of big data in the Internet of Vehicles,big data mining technology faces greater opportunities in the field of traffic safety.Based on the big data of the Internet of Vehicles,this paper preprocesses the collected truck trajectory data,dynamic data and early warning data,extracts the key data of truck travel,studies the risk characteristics of truck travel and its key influencing factors,in order to improve the safety of truck transportation process,Safety in the prevention and reduction of truck accidents.The main contents of the thesis are as follows:(1)Extraction,visualization and statistical analysis of early warning data of vehicle networking trucks.Through the cleaning and pre-processing of the original recorded data,932 Beijing-brand vehicles were extracted,including 716 vehicles with early warnings and 216 unwarned vehicles.Using programming technology to extract the target vehicle mileage,warning frequency,trajectory data and other information,map the extracted GPS trajectory data to achieve the visualization of the vehicle trajectory.From the vehicle trajectory distribution map,the "Beijing-Tianjin-Hebei" urban agglomeration The travel track has a high density.Using statistical analysis method,the distribution law of early warning frequency under different attribute characteristics is analyzed.The results show that vehicle type,link speed and visibility have significant influence on the frequency of vehicle warning.(2)Portrait of the risk characteristics of truck travel.First,with reference to historical driving characteristics such as driver mileage and warning frequency,cluster analysis method is used to classify it into different risk levels.The number of risk levels is determined by the "elbow" method.Then,the Bayes discriminant function is established for the vehicles under different risk levels to realize the risk identification of the vehicle.Finally,the decision and prediction of new data is realized by the discriminant function.The discriminant confusion matrix is calculated to obtain the recognition accuracy of the risk vehicle up to 95%,which indicates that the judgment model has a good interpretation degree,and the discriminant function can effectively identify the risk vehicle.(3)Analysis of key influencing factors of highway truck warning.With the acquisition of vehicle trajectory data and road network GIS base map,programming technology is used to achieve matching and extraction of typical high-speed road segments.Taking the "shuang yuan qiao" to "ma ju qiao" high-speed section of South 6th Ring Road as an example,variables are extracted from four aspects:driver factor,road segment factor,environmental factor and vehicle factor,and the Ordered probit model(ORP)is constructed to explore the influence of vehicles.The key factors of the warning,the model results show that the tractor is more prone to early warning than the ordinary truck,the worse the visibility is,the higher the possibility of warning,the higher the possibility of early warning in low temperature environment,early morning(24:00-06:00)and evening(18:00-24:00)is more likely to be early warning than other periods.The greater the average speed and acceleration,the higher the risk warning probability.The results of the two models show that the overall accuracy of SVM is better than that of ORP model(90.5%VS 85.6%).