| With the vigorous development of the transportation industry in our country,how to elevate traffic safety to a new level is the hot issue of extreme concern to people.For traffic safety development,more efforts need to be done to reduce the crash rates and the severity.However,due to the frequent unsafe driving behaviors,a potential safety problem for traffic crashes gradually are exposed,and traffic crashes occur frequently,which seriously hinders the development of traffic undertakings.From this,the study of the link between unsafe driving behaviors and traffic crashes will be a key to improving traffic safety.Based on the unsafe driving behaviors are the key factor related to crashes,relevant literature on unsafe driving behaviors and traffic crashes from domestic and abroad was reviewed and the shortage was summarized: First,the necessary traffic crash study data is inadequate,especially for the full story of the crash video.The traditional study of unsafe driving behaviors and traffic crashes exists deficiency based on the questionnaires or simulation experiments.However,the relationship between unsafe driving behavior and traffic crashes has not been carried out widely yet,as the crash data is difficult to obtain and the crash video is not fully disclosed.Second,the existing research lacks thorough and systematic studies of drivers’ unsafe driving behaviors and the chain of unsafe driving behaviors that precede crashes,which are often the key factors leading to traffic crashes.Third,at present,there is a finite amount of literature regarding unsafe driving behaviors and traffic crashes as two independent parts.Based on the spatio-temporal background,the correlation study between traffic crashes and traffic violation behaviors is very limited.Fourth,there is a lack of research on how to make targeted traffic crashes prediction based on the easily obtained unsafe driving behavior data.This study established the analysis model to explore the relationship between unsafe driving behavior and traffic crashes with more refined data.On the one hand,it was advantageous to study unsafe driving behaviors through crash video,which could help to focus on the latent unsafe driving behavior matching collisions on urban expressways,not only focusing on high-risk unsafe driving behaviors,but also the chain combination among unsafe driving behaviors.Partial crashes can be avoided if the unsafe behavior is corrected in a timely manner across risk stages.On the other hand,this study intended to explore the relationship between traffic violations and traffic fatal crashes.The model was constructed on the basis of risky factors leading to crashes,and the spatial aggregation,time aggregation,and the space-time aggregation interaction was considered among factors.With easy access to traffic violations data,the zero-value problem of traffic fatal crashes was effectively forecasted,which helped to master the development law of traffic fatal crashes.More specifically,this study included the following contents:First,the unsafe driving behavior data and traffic crash data used in the research were elaborated.In addition,the collected data were analyzed with the help of statistics,machine learning,geographic information technology and other methods.Second,fourteen unsafe driving behaviors were identified through real crash videos analyzed for crashes on urban expressways.A topology diagram of unsafe behaviors was then constructed to investigate the relationship between drivers’ unsafe behaviors and crashes.Correlation inference,prediction inference,and diagnostic inference were performed for crashes on urban expressways.An approach including the process analysis has been proposed to uncover chains of high-risk unsafe behaviors.Third,this study explored the relationship between various types of traffic unsafe driving behaviors and traffic crash risky stage on urban expressways.The correlations between unsafe driving behaviors and traffic crash types were identified.The analyses conducted with the survival analysis methodology yielded a clear picture of the main behavior patterns causing crashes on urban expressways.The time of the traffic crash risky stage on urban expressways was quantified through the reasoning of key unsafe behaviors for different types of crashes.Fourth,according to the characteristics of the data of traffic violations and traffic fatal crashes.The multilevel analysis model was constructed based to identify the relationship between traffic violations and traffic fatal crashes considering the spatial aggregation,time aggregation,and space-time aggregation interaction.Finally,based on the correlation between traffic violation behaviors and traffic fatal crashes in the temporal and spatial background,by the traffic violation data,neural sequence network,long and short-term memory neural network,time series model were used to build crash forecasting model,which mastered the occurrence rules of traffic fatal crashes.This study compared the performance and impact of the prediction model under different spatio-temporal conditions.Proposed predictions for fatal traffic crashes were identified.This study makes some contributions to the relationship between traffic crashes and unsafe driving behaviors.The study can help to prevent traffic crashes from the source and to grasp the traffic crash content.The findings could help drivers correct their unsafe behaviors,so that the crash occurrence and crash rate can be reduced.The results of this study have important theoretical value and practical significance for upgrading traffic safety level. |