Analysis And Prediction Of Freeway Accident Risk Based On Interaction Between Weather Condition And Traffic Flow

Road traffic accidents are now the fifth leading cause of death worldwide,with more serious accidents and more casualties on freeways.Traditional risk management for freeway is always passive and static,focusing on the concept of “post-emergency management”,which cannot meet the increasing needs for modern freeway management.In recent years,the rapid development of data collection and analysis,as well as the research and development of traffic accident early warning system,transferred freeway risk management from "post-emergency management" to "proactive prevention and control".The basic concept of proactive management for freeway risk is described as follows.Based on the analysis for high traffic flow characteristics,weather conditions relevant to freeway traffic accident,data mining methods can be applied to identify risk factors affecting freeway safety,to judge freeway safety status,as well as to get an accurate prediction for whether an accident will occur under current freeway operating conditions.With the analysis and prediction result,administrative department could carry out targeted freeway early warning and emergency arrangement in advance,improve freeway safety management and control,and reduce the number of traffic accidents as well as the severity of accidents.At present,freeway and risk management in our country is still in its infancy,with an increasing need for a series of scientific and practical methods.The present study takes freeway safety risk as the research subject,starts from the interaction of freeway traffic flow characteristics and weather conditions,utilizing the theory of risk analysis,and conducts a study on freeway risk proactive management from risk identification,risk assessment and risk prediction.The research could provide decision-making support for administrative department to formulate risk prevention and control measures.The present study mainly covers the following three aspects:(1)Method study on freeway accident risk identificationFirstly,the association rule mining method is applied to identify the risk factors that affect the safe operation of freeway.Through collecting,sorting and analyzing the statistical data of freeway traffic accidents,the risk factors affecting the freeway safety are found to be multi-level,multi-dimensional and correlated.Thus,the freeway risk factor identification method of multi-dimensional and mult-ilayer AHP weighted Top-k association rule mining algorithm based on improved Apriori algorithm is utilized.This method surpasses the traditional statistical analysis method on its capability to figure out the correlation among different influence factors,as well as its capability to avoid the need to set the minimum support threshold in the process of mining,which may lead to missing important association rules,and thus achieves a deep analysis of the risk factors that affect the safe operation of the freeway.(2)Freeway safety state classification and assessment on the basis of the interaction between weather conditions and traffic flowSecondly,the freeway safety risk state is evaluated.For the analysis of the quantitative relationship among freeway operation safety,traffic flow dynamic characteristics and weather conditions,a "case-control" sample structure data matching method is applied to extract of upstream and downstream traffic flow state data and corresponding high-precision weather data before a freeway traffic accident takes place at a certain accident location,which,meanwhile,can avoid the influence of other confounding factors.Based on these efforts,the composite index that can represent the state of traffic safety is calculated,and the important index with the highest variable importance is screened out utilizing the random forest algorithm,thus solving the problem of dimensional disaster in the later calculation process.Then,in accordance with the mixed data characteristic that the evaluation index contains both numeric data and subtype data,a new freeway safety state classification method is proposed based on fuzzy algorithm of k-prototypes.Then,a Bayesian logistic regression method is applied to evaluate the influence of different safety states on accident risk.The study shows that the freeway safety state could be classified into 6 categories considering both traffic flow characteristics and weather conditions,while the safety state could be classified into 5 categories considering solely traffic flow characteristics,which indicates that the classification is more accurate considering both weather conditions and traffic flow characteristics.(3)Freeway accident risk prediction on the basis of the interaction between weather conditions and traffic flowFinally,the freeway safety risks are predicted.The interaction between weather conditions and traffic flow characteristics is quantitatively analyzed,and it is proved that the inclusion of weather conditions in freeway safety risk assessment can improve the accuracy of freeway risk identification.Then,a freeway accident risk prediction model based on learning rate adaptive stochastic gradient enhancement algorithm is proposed to solve the correlation between weather conditions and traffic flow characteristics in the prediction model.The research shows that the freeway safety risk evaluation and predition model considering both weather conditions and traffic flow characteristics has the highest accuracy,and the prediction accuracy is relatively high with massing data and outlier.