Research On Road Traffic Safety Risk Based On Traffic Violation Events

Preventing and reducing traffic crashes have become an important problem for sustainable development of cities in China.Although the government and management departments have taken some measures in law,management,publicity and education,the traffic safety problems are still rigorous.Therefore,proposing a “pre-prevention” method to change the traditional“passive” method is very important for traffic safety.Based on the active management of road traffic safety,this thesis chooses traffic violations as an identification factor of traffic safety risk,and the structure of thesis is organized as three parts: characteristic research,mechanism research and application research.Specifically,the main contents of this thesis can be divided into the following areas:(1)The study proposed a spatio-temporal kernel density estimation(STKDE)model to reveal the spatial and temporal patterns of traffic violation.The STKDE model combined a bivariate spatial kernel function with a univariate temporal kernel function to estimate the density of violation events,and integrated spatio-temporal statistics and three-dimensional visualization techniques.The Plug-in method was selected for optimal bandwidth selector,and the space-time cube method was established to achieve the best visualization result.(2)The study analyzed and predicted the spatiotemporal evolution of traffic violations.The geographically and temporally weighted regression(GTWR)model,integrating spatiotemporal proximity and spatio-temporal weight matrix,was applied to analyze the spatiotemporal heterogeneity and spatiotemporal dependence of traffic violations.The GetisOrd Gi* model was used to calculate the Z scores and P values of different violation positions.The Mann-Kendall trend test method was used to predict the hot and cold points of violation positions,and determine the classification result for the evolution of traffic violations.(3)The study quantified the relationship between traffic violations and various characteristic variables.The characteristic variables of traffic violations can be divided into time factors,spatial factors,traffic factors and weather factors.A binomial Logit model was proposed to explore the influence of characteristic variables on the probability of traffic violations.Also,the relationship of three types of traffic violations "red light pressure line","vehicle illegal parking" and "running red light" were studied by the mixed Logit model.The results showed the presence of regularity between traffic violations and various characteristic variables,and confirmed the validity of traffic violation for traffic safety risk.(4)The study proposed a data mining model to find the association rules between traffic violations and traffic crashes.In order to solve the problems of small sample size and subjectivity of traditional traffic crash data,the traffic alarm call data was selected as the source of traffic crash data.A traffic crash information extraction method based on dependency syntax analysis was established to obtain the different types of traffic crash.Combined with traffic crashes and traffic violation data under the condition of time and space constraints,it calculated the association rules between traffic crashes and violations based on FP-growth algorithm.(5)The study proposed a road traffic safety risk prediction model under the continuous data environment.This model developed the traffic safety risk prediction techniques based on the improved random forest(IRF)algorithm,and focused on solving the problems of unbalanced data and feature variable selection.A three-in-one road traffic safety risk management and control framework was established,which combines "monitoring","discovery" and "solution".Also,some countermeasures were recommended to prevent and control traffic violations.The work in the thesis can help us better understand the application of traffic violation data in road traffic safety risk.It is also provides the technical support and scientific basis for traffic safety improvement.