Apriori algorithm code.

The real-time monitoring on the risk status of the vehicle and its driver can provide the assistance for the early detection and blocking control of single-vehicle accidents. However, complex risk coupling relationship is one of the main features of single-vehicle accidents with high mortality rate. On the basis of investigating the coupling effect among multi-risk factors and establishing a safety management database throughout the life cycle of vehicles, single-vehicle driving risk network (SVDRN) with a three-level threshold was developed, and its topology features were analyzed to assessment the importance of nodes. To avoid the one-sidedness of single indicator, the multi-attribute comprehensive evaluation model was applied to measure the comprehensive effect of characteristic indicators for nodes importance. A algorithm for real-time monitoring of vehicle driving risk status was proposed to identify key risk chains. The result revealed that improper operation, speeding, loss of vehicle control and inefficient driver management were the sequence of top four risk factors in the comprehensive evaluation result of nodes importance (mean value = 0.185, SD = 0.119). There were minor differences of 0.017 in the node importance among environmental factors, among which non-standard road alignment had the larger value. The improper operation and non-standard road alignment were the highest combination correlation of factors affecting road safety, with the support of 51.81% and the confidence of 69.35%. This identification algorithm of key risk chains that combines node importance and its risk state threshold can effectively determine the high-frequency risk transmission paths and risk factors through multi-vehicle test, providing a basis for centralization management of transport enterprises.

对车辆及驾驶员的风险状态开展实时监测，可为单车事故的早期发现与阻断管控提供辅助支撑。然而，高致死率单车事故的核心特征之一，便是复杂的风险耦合关联。本研究在探究多风险因素间耦合效应、构建车辆全生命周期安全管理数据库的基础上，搭建了具备三级阈值的单车行驶风险网络（single-vehicle driving risk network, SVDRN），并通过分析其拓扑特征以评估节点重要性。为规避单一指标评价的片面性，本研究采用多属性综合评价模型，测算节点重要性特征指标的综合效应。本研究提出一种车辆行驶风险状态实时监测算法，用于识别关键风险链。研究结果显示，在节点重要性综合评价结果中，排名前四的风险因素依次为操作不当、超速行驶、车辆失控以及驾驶员管理效能低下，其均值为0.185，标准差为0.119。环境因素对应的节点重要性差异极小，组间差值仅为0.017，其中非标准道路线形的节点重要性分值更高。操作不当与非标准道路线形是影响道路安全的因素中组合关联度最高的因子对，其支持度为51.81%，置信度为69.35%。这种融合节点重要性与风险状态阈值的关键风险链识别算法，经多车辆测试可有效甄别高频风险传导路径与风险因素，可为运输企业的集约化管理提供决策依据。