Research On Influencing Factors Analysis And Severity Forecasting Of Road Traffic Accident In China

Transportation is the basis of urban and social development,and it is also a ropeway system establishing ties with all social activities.Recently,with great economic and social progress,China has promoted urbanization comprehensively,increasing motorization across the country,which results in a continuous growth of road traffic demand and motor vehicle ownership.However,the improvement of transportation infrastructural construction and road safety management and control obviously lags behind the development of traffic demand,which in turn boosts pressure on traffic safety and remains the number of road traffic accidents at a high level.Nowadays,road safety has become a widespread social problem,significantly influencing personal safety and the property security of people.Besides,the coordinated development of society and the economy road safety is restricted by road safety issues.Therefore,how to carry out scientific and effective road safety control has become the focus of traffic safety research.The elements involved in road traffic accidents are numerous and complex,with fortuity and ambiguity.Nevertheless,existing research shows that under the appearance of fortuity,the occurrence and severity of traffic accidents will be impacted by human,vehicles,roads,environment,and etc.in a regular pattern.Recognizing the inherent and objective rule of road traffic accidents is of great practical significance to avoiding traffic accidents,decreasing the number of injuries and deaths related to traffic accidents,and lowering the severity of road traffic accidents.This paper researches the relevant influencing factors of road traffic accidents,applying Factor Analysis to rank the factors of road traffic accidents according to their importance and explore the main factors of traffic accidents,and utilizing the random forest algorithm to build a road traffic accident severity prediction model to provide suggestions for road safety.The research mainly includes the following sections:Firstly,this paper selects 18,858 real road traffic accident data in China,establishes a factor analysis model based on factors such as people,vehicles,roads and the environment,obtains the public factors by the built model,and then calculate the importance degree of each factor by the improved weight model.The study found that age classification,weather,type of people,road surface condition,residence type,roadside protection infrastructure type,and road physical isolation were the most significant influencing factors in road traffic accidents in China;while the third-party liability insurance,road condition,the type of road sections and intersections,pavement structure,cross-section location and type of drivers ’ licenses have the least significant impact on the severity of road traffic accidents.Secondly,on the basis of the public factors extracted from the factor analysis process and the weight values,which are calculated by the improved weight model,of influencing factors on the public factors to which they belong,12 public factors are converted from the 33 features included in the original data,a public factor data set is established,and a FA-RF-based road traffic accident severity prediction model is constructed by the random forest algorithm.Finally,this paper uses the FA-RF-based road traffic accident severity prediction model to predict two-category accidents and three-category accidents,respectively,and compares the prediction results with those of the random forest single model.According to the model results,in the two-category accident prediction,the accuracy of the random forest single model is 94.22%,and the accuracy of the FA-RF-based fusion model is 96.29%,showing a 2.07% growth;in the three-category accident prediction,the accuracy of the random forest single model is 67.60%,and the accuracy of the FA-RF-based fusion model is 75.87%,increasing 8.27%.Demonstrated by the comparison,whether it is a two-category accident prediction or a three-category accident prediction,the FA-RF-based fusion model predicts more accurate than the random forest single model,which verifies the feasibility of the fusion model based on FA-RF.By comprehensively comparing the classification results of the two-category accident prediction and the three-category accident prediction,this paper also finds that the main reason for the accuracy decline of the FA-RF-based fusion model between the three-category accident prediction and two-category accident prediction is that the model has a large degree of misjudgment for injuries and fatalities.In other words,the current collection of road traffic accident information in China is still insufficient,and more factors related to casualties should be considered.