基于贝叶斯网络的城市平面交叉口交通事故分析

应用贝叶斯网络对城市平面交叉口交通事故进行了分析。以3 584起交通事故数据为分析依据,基于专家知识和数据融合方法建立了城市平面交叉口交通事故分析的贝叶斯网络结构,利用服从Drichlet分布的贝叶斯方法对贝叶斯网络进行了参数学习。结合网络模型,应用联合树引擎算法推断了在车辆类型、交叉口类型、交叉口控制方式和交通参与者等因素的影响下平面交叉口交通事故类型的变化。研究结果表明,在城市平面交叉口中,由自行车导致的正面碰撞事故的概率最大,为22.83%,由于交通参与者转向不当引起的侧面碰撞的概率为23.44%,同时也易导致刮擦事故的发生;交通参与者的感知判断失误导致尾随碰撞事故的概率为23.62%。

Analysis of Urban Intersection Traffic Accidents Based on Bayesian Network

A Bayesian network is developed to analyze the urban intersection traffic accidents.Based on 3,584 recorded urban intersection traffic accidents,Bayesian network structure is formed with references to expert knowledge and data fusion method.The process of parameter learning is completed based on Dirichlet distribution.The influences of some factors on the intersection accidents,such as the vehicle type,intersection type,control style and traffic participant,are computed using the junction tree engine built on Bayesian network and recorded accident data.The results indicate that,the probability of direct crash caused by bicycle is 22.38%.In comparison,steering failure and judgment failure make the largest probability of side-crashes and rear-end crashes,at 23.44% and 23.62%,respectively.