电动自行车与机动车事故严重性影响因素分析

为减少电动自行车与机动车事故造成的损失,定量剖析不同因素对事故严重程度的差异性影响至关重要。基于上虞区2018年10304起电动自行车与机动车事故,分析该类事故的严重性分布情况和时间与空间分布特性。以事故严重性为因变量,将其有序分为未受伤、轻伤及重伤事故3类,从时间、空间、道路、环境、骑行者及车型6个方面,选取17个事故严重性潜在影响因素,采用多项Logit模型、有序Logit模型、广义有序Logit模型及偏比例优势模型进行拟合度对比分析,并以最佳模型(偏比例优势模型)和边际效应,量化分析各因素对事故严重性影响的显著性与差异性。结果表明,除节日类型、车道限速、车流相交角及温度对事故严重性影响不显著外,其余13个因素都有显著影响,事故时间、光线亮度、骑者性别、骑者年龄、车辆类型及电动自行车类型违反平行线假设;对该类事故严重性影响最大的前4个因素为电动自行车类型、机动车类型、骑行者年龄与性别,其边际效应绝对值的最大值均超过61%,事故区位、道路类型及光线亮度的影响较大(20%~30%),事故时间和风力等级影响较小(10%~20%),而季节、事故位置、慢行干扰度和天气状况的影响最小(≤8%)。基于各因素的差异性影响,为交通管理部门提出了有效改善建议。

Analysis on Contributing Factors Influencing Severity of E-Bicycle and Motor Vehicle Crashes

It is critical to quantify the differential influence of different factors on the severity of e-bicycle and motorvehicle crashes in order to mitigate the losses caused by these crashes. Based on the data of 10304 e-bicycle and motorvehicle crashes in Shangyu in 2018, the severity distribution and spatiotemporal characteristics of the crashes wereanalyzed. As the dependent variable, the crash injury severity levels were coded into three classifications: uninjured,minor injury, and serious injury crashes. A total of 17 contributing factors were selected from six perspectives, i.e.,time, space, road, environment, cyclist, and vehicle type. The multinomial Logit model, the ordered Logit model, thegeneralized ordered Logit model, and the partial proportional odds model were applied to explore the potentialinfluencing factors by comparing the goodness of fit. The optimal model, i.e., the partial proportional odds model, aswell as the marginal effects was carried out to quantitatively analyze the significance of the contributing factors in thecrashes. The results evidenced that, the day type, the speed limit of lanes, intersection angle of traffic flow, andtemperature have no significant influence on the crash severity, while the other 13 factors have significant influence.The factors, including crash time, lighting condition, rider gender, rider age and vehicle type, do not comply with theparallel-lines assumption in the crashes. The top four factors with the greatest impacts were e-bicycle type, vehicletype, rider age, and gender. The maximum absolute marginal values all exceeded 61% . Crash area, road type, andlighting condition had moderate impacts (20% ~30% ), crash time and wind level had less impact (10% ~20% ), andseason, crash location, non-motorized traffic interference, and the weather had minor effects (≤8% ). Based on thedifferences in the influencing factors, effective improvement suggestions had been put forward for the trafficmanagement department.