Analysis Of Influencing Factors Of Road Transportation Accidents Of Dangerous Goods Based On BTM Model

With the rapid development of chemical industry,the number of dangerous goods manufactured is increasing,and the demand for dangerous goods is also increasing.Therefore,dangerous goods transportation accidents occur from time to time in China,with serious consequences and a threat to road traffic safety.However,most of the existing studies focus on surface data statistics.In order to improve this situation,this paper analyzes the accident text in the field of road transportation of dangerous goods,The study is of great significance to improve the safety level of road transportation of dangerous goods and reduce the incidence of accidents.The main research contents of this paper are as follows:(1)Design data acquisition module and urban weather service system,establish dangerous goods transportation accident database,collect 1247 data from 2016 to 2021,extract text data in the database and clean accident data.Based on the corpus,this paper establishes a special dictionary for dangerous goods,uses Jieba database for text word segmentation,extracts the text features in the corpus,screens the text keywords,and analyzes the risk of dangerous goods transportation accidents.(2)The PMI point mutual information method is used to mine frequent co-occurrence words in the accident text,visualize the co-occurrence word mining results,carry out topic mining based on the improved BTM model based on PMI,determine the optimal number range of topics by using topic consistency and confusion,determine the optimal number of topics by pyldavis topic visualization,and finally extract 7 cause topics and 3 accident mode topics,and divide the topics into active cause There are three types of passive causes and accident modes,and active causes,including improper driving behavior,weak safety awareness,driving qualification and inadequate management.These four types of topics account for 40.82%.Passive causes,including vehicle hardware damage,bad weather and external traffic environment,accounted for32.43%.The accident modes include vehicle spontaneous combustion,rear end collision and unilateral accident,accounting for 26.75%.(3)The accident causing ontology is abstracted into seven subclasses,and the event causing ontology and the event causing network are used as two subclasses.The event causing ontology is abstracted into two subclasses,including the event causing attribute and the event causing attribute.The accident probability reasoning of ontology Bayesian model is carried out to explore the probability of each parent node in the case of different levels of accidents;Secondly,taking seven causative themes as parent nodes and accident types as target nodes,the influence degree of abstract causative themes on accident types is speculated.The reasoning results show that in the three types of accidents,the sum of posterior probabilities of each causative theme is greater than 1,indicating that each theme does not act alone,and each theme node is coupled and interacted with each other,resulting in accidents;Finally,sensitivity analysis is carried out,taking seven cause subject nodes as the target nodes,so the whole Bayesian network is divided into seven sub Bayesian networks to analyze the most sensitive influencing factors under each cause subject in case of minor accident,general accident and serious accident,so as to provide constructive suggestions for road transportation of dangerous goods.