基于APRIORI-TAN的交通事故伤害分析与预测  

Analysis and prediction of traffic accident injury based on APRIORI-TAN

作　　者：韩天园 吕凯光 许江超 李旋 乔洁[1] HAN Tianyuan;LYU Kaiguang;XU Jiangchao;LI Xuan;QIAO Jie(School of Automobile,Chang’an University,Xi’an Shaanxi 710064,China)

机构地区：[1]长安大学汽车学院,陕西西安710064

出　　处：《中国安全生产科学技术》2021年第8期50-56,共7页Journal of Safety Science and Technology

基　　金：陕西省重点研发计划项目(2020ZDLGY16-08);教育部人文社会科学青年基金项目(18YJCZH110)。

摘　　要：为探究道路交通事故因素和事故伤害的相关性,以2467起涉及人员伤亡的交通事故为数据集,运用Apriori算法分别挖掘事故伤害关联规则,并结合社会网络分析的可视化和核心-边缘分析构建受伤事故和死亡事故的关联规则网络。结果表明:事故伤害程度与事故时间、道路条件和交通环境等因素关系紧密,尤其死亡事故与碰撞固定物、人行横道事故、高速公路、高速道路、非市区、酒驾和超速存在高相关性。基于树型贝叶斯网络(TAN)构建事故伤害程度的预测模型,预测结果准确率可达87.56%。In order to explore the correlation between road traffic accident factors and accident injuries,taking 2467 traffic accidents involving casualties as the data set,the Apriori algorithm was used to mine the association rules of accident injury respectively,and the association rule network of injury accidents and fatal accidents was constructed combined with the visualization of social network analysis and core-edge analysis.The results showed that the degree of accident injury was closely related to the factors such as accident time,road conditions and traffic environment,etc.In particular,there was a high correlation between the fatal accidents and collision with fixed objects,crosswalk accidents,expressways,high-speed roads,non-urban areas,drunk driving and speeding.Finally,a prediction model of accident injury degree was constructed based on the tree Bayesian network(TAN),and the accuracy of prediction results could reach 87.56%.

关 键 词：交通安全 事故伤害 关联规则 社会网络分析 树型贝叶斯网络 伤害预测 

分 类 号：X928.03[环境科学与工程—安全科学]