Prediction Of Accident Severity Of Two-wheeled Vehicles Based On Machine Learning

Road traffic is the foundation of social development and progress,and it is an important guarantee for people to become rich and strong.In recent decades,China’s economy has developed rapidly,and the same is true in the field of transportation.In the process of promoting urbanization in an all-round way in China,not only the total mileage of roads is increasing,but also the number of motor vehicles and non-motor vehicles is increasing rapidly.Two-wheeled vehicles,including electric two-wheeled vehicles and two-wheeled motorcycles,have developed rapidly in recent ten years,but they have also led to a series of traffic and social problems.With the number of two-wheeled vehicle traffic accidents in China constantly creating a record high and the injuries caused by accidents becoming more and more serious,in order to reduce the occurrence and severity of two-wheeled vehicle traffic accidents,it is even more urgent to study the occurrence law of two-wheeled vehicle accidents and explore the factors affecting the severity of accidents.Firstly,the statistical characteristics of accident data of two-wheeled vehicles are analyzed to study the change of severity of traffic accidents of two-wheeled vehicles under different conditions.Understand the general characteristics and development trend of traffic accidents of two-wheeled vehicles,and quantitatively understand the nature and potential laws of traffic accidents of two-wheeled vehicles from a macro perspective.Secondly,factor analysis is used to classify feature attributes and extract common factors.By analyzing the correlation between feature attributes and common factors,and determining the subordinate relationship,the index framework of the severity of two-wheeled vehicle traffic accidents is constructed.Thirdly,according to the factor analysis method and the improved weight model,the weight value of each feature attribute is calculated,and the three-tier index weight value between the feature attribute and the common factor and between the two-wheeled vehicle accident severity is formed.According to the weight value of feature attributes,the important factors affecting the severity of two-wheeled vehicle accidents are determined,and the causes are qualitatively analyzed,and targeted improvement suggestions are put forward.Finally,a random forest model is established,and the original data set and the common factor data set are taken as the input conditions of the model,and a single random forest model and a random forest combination model are obtained.Similarly,two different data sets are used as input conditions of other classification models to form a classification combination model.According to the prediction results of the model and the evaluation index system of the model,the model is evaluated in all directions.According to the optimal model,the correlation between key feature attributes and the severity of traffic accidents of two-wheeled vehicles is further studied,and the optimal value of related feature attributes is determined.The results show that the factor analysis model established in this paper can classify the feature attributes and calculate their weights.According to the sorting results of feature attributes,the important feature attributes are determined,and targeted improvement measures are put forward,which can provide theoretical basis for urban road planning and the formulation of two-wheeled vehicle restriction policies.In addition,the accuracy of the combination of random forest model and factor analysis model is improved by 4.4% compared with single random forest model,and the combination effect of random forest model and factor analysis model is better than that of other classification models and factor analysis models.