基于贝叶斯网络的网约车交通事故致因机理分析

为了缓解网约车交通事故，以网约车驾驶员为研究对象，探究网约车交通事故致因机理.通过调查问卷收集了2 458名网约车驾驶员个体属性、工作强度、工作压力、不良驾驶行为和交通事故经历的相关信息.对数据进行分类处理后，通过贝叶斯网络建立网约车事故频率预测模型.基于十折交叉验证法，使用混淆矩阵与接收者操作特征曲线校验模型精度.结果表明，模型预测能力较好，模型分析了11种与事故频率直接相关的影响因素，识别了16类导致高频率事故发生概率增加的不利状态，且发现多种不利状态组合对事故频率产生的非线性扩增效应和叠加效应.研究结论有助于管理部门制定相应预防对策以减少网约车交通事故频率.

Analyzing the Causes of Traffic Accidents of Online Ride-Hailing Cars Using the Bayesian Network

To mitigate the traffic accidents of online ride-hailing cars， the accident causes were studied by taking the ride-hailing drivers as the research object. A self-reported questionnaire survey was conducted to collect information about self characteristics， work intensity， work stress， risky driving behaviors， and accident history for 2458 ride-hailing drivers. After sorting of data， the Bayesian network method was used to establish the prediction model of accident frequency. The accuracy of the model was calibrated using the confusion matrix and receiver operating characteristic curve based on the ten-fold cross-validation. The results show that the model has a good prediction ability. The model found 11 influencing factors directly related to accident frequency and identified 16 categories of unfavorable states leading to an increased probability of high-frequency accidents. Also， nonlinear amplification and superposition effects of the combination of multiple unfavorable states on accident frequency were confirmed. The conclusions of the study help the management department to make prevention countermeasures to reduce the accident frequency of online ride-hailing cars.