China's private car fleet reaches 262 million by 2021. With the increase in the number of cars, traffic accidents have become a major social safety issue1. Every year, about 1.2 million people die in traffic accidents worldwide, and tens of millions more are injured or disabled in traffic accidents2,3. The number of traffic accidents and their tragic consequences reveal the need to study the causes of traffic accidents. In order to make a breakthrough in the depth study of road traffic accidents, each country has established a vehicle accident depth study database. For example, the German Federal Institute of Transport (BASt) and the US National Highway Traffic Safety Administration (NHTSA) have established in-depth road traffic accident research systems and databases respectively. China established the China Vehicle Accident In-Depth Investigation System NAIS (National Automobile Accident In-Depth Investigation System) in 2011 by the Defective Products Management Center of the General Administration of Quality Supervision, Inspection and Quarantine of China. In 2020, there were 244674 traffic accidents in China, with 1349 pedestrians killed in accidents and direct economic losses of nearly 30 million yuan. According to the analysis of the NAIS database on the in-depth survey data of vehicle accidents in 2020, the number of pedestrian-vehicle traffic accidents ranked second only to automobiles and two/three-wheelers among all types of traffic accidents. Pedestrians have received a great deal of attention as a vulnerable group of road users4. Especially with the progress of intelligent driving technology, the research on the causes of pedestrian-vehicle traffic accidents can not only provide a reference for traffic technology, but also provide a basis for government legislation, etc.
Regarding the causal factors of traffic accidents, on the study of human factors, Topolšek5 obtained that Wrong-way driving is one of the causes of accidents on highways; Komada6 found drowsiness as a significant factor in human error leading to traffic accidents; Li7 identified not giving way, illegal reversing, illegal parking, drunk driving, and speeding as high-risk behaviors; Wang8 Analyzed the severity of injuries in traffic accidents caused by driver fatigue. On the study of vehicle factors, Vranješ9 obtained that the main vehicle factors leading to traffic accidents are the malfunctioning of lights or light signal devices. On the study of road and environmental factors, Kurdin10 analyzed the influence of geometry and road environment on road accidents in Kecamatan Abeli; Liu11 found that severe weather has a significant impact on the consequences of high-speed accidents. In addition, Angin and Ali12 found that negligent driving and speeding are the main causes of road traffic accidents through a study of people, vehicle, road and the environment; Hauer13 studied the relationship between average annual daily traffic volume, number of commercial lanes, and speed limits and the probability of traffic accidents; Hou14 conducted a detailed inference analysis based on traffic characteristics, highway geometry, pavement conditions, and weather conditions to derive the main causal factors of highway vehicle crashes; Li and Liu15 identified time of day, weather, number of patrol vehicles and surveillance, and age and attributes of the accident driver as some of the factors contributing to high speed traffic accidents; Traffic accidents are the result of a combination of factors, human, vehicle, road and the environment16. In this paper, a total of 29 basic pedestrian-vehicle traffic accident causative factors, including drunk driving and pedestrians crossing the traffic lane illegally, are also obtained from three categories: people, vehicle, road and environment.
Currently, there are various methods used to study the causes of traffic accidents. For example, Keay and Simmonds17 designed a regression model for the effect of weather on traffic; Yu18 established a highway collision model using Poisson model and Bayesian inference method. Selvasofia and Arulraj19 used GIS for traffic accident analysis; Zamzuri20 used Bayesian network and HC algorithm and Tabu algorithm to explore the causes of traffic accidents in Malaysia; Zhang21 used text processing technology based on LDA Topic Model to analyze traffic accidents to obtain the most dominant factors of traffic accidents; Deng22 proposed a causality analysis model for traffic accidents using a hybrid AHP and Apriori-Gentic based algorithm to mine accident causes. Bayesian network is a good way to model causality, and Bayesian network can effectively form probabilistic models for efficient inference and learning23. Fault tree analysis is an important analysis method for system reliability and safety analysis24. In this paper, we combine fault tree and Bayesian network methods to build a model of pedestrian-vehicle traffic accidents, which can more accurately identify the factors affecting pedestrian-vehicle traffic accidents. We first construct a fault tree model for pedestrian-vehicle traffic accidents, and then transform the fault tree model into a Bayesian network model, which can generate the relationship structure between variables. Bayesian network can quantify the relationships between variables as probabilities, and using these probability values can make it relatively easy to explore the relationships between variables and make inference about pedestrian-vehicle traffic accidents25.
The main objective of this paper is to analyze the main causal factors leading to pedestrian-vehicle traffic accidents using fault tree and Bayesian network models. Provide guidance for the traffic safety department to formulate corresponding preventive measures, reduce the number of traffic accidents of people and vehicles, and reduce the number of deaths and injuries.