Construction And Application Of Fatal Traffic Accident Database Based On FARS

For in-depth investigation of traffic accidents,China established China In-Depth Accident Study(CIDAS)in 2011 and National Automobile Accident In-Depth Investigation System(NAIS)in 2012 based on Germany In-Depth Accident Study(GIDAS).However,both databases were mainly from the perspective of vehicles,and were dedicated to discovering automobile defects,ignoring pedestrians and other factors.The data elements are also complex and not specifically targeted at fatal traffic accidents.China has a large amount of fatal traffic accidents while there were few related studies.The value of data directly borrowed from other countries was low.Therefore,this study explores the establishment of fatal traffic accidents database in China on the basis of Fatality Analysis Reporting System(FARS)in the United States and some of its application demonstrations.The FARS was processed as follows to build the database structure quickly.Firstly,simplified the data relationship of the FARS,and only kept three data files of accident,vehicle and person,and the remaining data files were attributed to the three files according to the description objects.Secondly,the new data were filled into three tables of accident,vehicle and person by means of inheritance,addition,deletion,modification.Based on these three tables,information such as traffic accidents and traffic accident scene maps that were not available in the FARS but had practical research value were added.The duplicated,terminated,domestic vacancies data elements were deleted.Data information of traffic accidents in China has been improved and converted to meet the requirements.The resulting accident,vehicle and personnel table was in new database data form,and each form was included in the data file of the same name.The data forms contained in the database established above were accident form,vehicle form and person form.The content entered in the accident form was natural environment information,road information,accident information and so on,corresponding to multiple vehicle data forms.The vehicle form recorded the vehicle operation information and driver information before accident,and the vehicle information after accident,corresponding to multiple person information forms.The person form recorded information about all occupants,including personnel basic information,personnel status at the time of the accident,personnel injuries,etc.All non-motorized personnel were occupants of "vehicle 0".Python and Django were used to establish database framework in the form of a web page,which was implemented and ran on PyCharm.171 major traffic accidents with more than 10 deaths nationwide from 2009 to 2017 and 355 domestic fatal traffic accidents from 2012 to 2016 were recorded according to the definition of fatal traffic accident of FARS.A fatal traffic accident data entry system has been preliminarily established in line with China's national conditions,which paves the way for the follow-up database application research.A total of 158 major traffic accidents with complete information were screened out from the database.The neural network model was used to explore the relationship between traffic accident scene information and vehicle occupant casualty rate.According to the chronological order,the first 120 cases were training set,the last 38 cases were test set,and the neural network model was continuously debugged.Finally,a neural network model with an accuracy of 0.8158 was obtained.Considering the limitation of data volume and data information,the results obtained by the model were considered to be credible,and the fatal injury status of vehicle occupants after the accident can be quickly assessed.A comparative analysis of 199 cases of fatal pedestrian traffic accidents in Chongqing and fatal pedestrian traffic accidents in FARS showed that there were differences in structure and proportion.Logistic regression was used for further modeling and analyzing the risk of fatal injuries suffered by pedestrians in both countries under age,light condition and vehicle type.The highest risk of pedestrians in the United States are daylight conditions,children and adolescents,and motorcycle models while the highest risk of pedestrians in China is dawn,middle-aged and van models.Therefore,China should consider these differences when using the foreign data and experience.Based on FARS,this study attempted to develop a database of fatal traffic accidents in China.It has been tested and operated well and meets the characteristics of traffic accidents in China.Up to now,526 cases were recorded,the evaluation of vehicle occupant injury and pedestrian safety was basically achieved,proving that it was feasible to study fatal traffic accidents in the form of database.It can be useful for the in-depth investigation of traffic accidents and the formulation of various rules and regulations in China.