Taking traffic accident data of Northern California in recent ten years as the research data set, the Bayesian network structure of causal inference of traffic accidents is established based on expert knowledge and gradient descent algorithm. Twenty-five variables in the data set are selected as the parameters of the Bayesian network for learning. Combined with the Bayesian network structure, the probability distribution of traffic accident casualties under the influence of ethnic distribution, population number, traffic volume, roadrelated conditions, and other factors is inferred, and the causal relationship between 25 traffic factors and traffic accident casualties is analyzed. The results show that under the influence of ethnic distribution and population, the probability of death is 0.169 and injury is 0.164; Under the influence of traffic volume, the probability of death is 0.19 and injury is 0.162; Under the influence of road-related conditions, the probability of death is 0.165 and injury is 0.172. The Bayesian network model has high accuracy, and the error rate accuracy is only 8.889%. Related research can provide a theoretical basis for traffic measures intervention.
The casual inference of road traffic accidents based on the Bayesian network optimization
01.04.2023
582041 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch