基于二元Logistic模型的城市道路交通事故严重程度分析

为从多维度精准剖析影响城市道路交通事故严重程度的因素，选取了我国某城市2018—2020年交通事故数据库中的4 587条数据作为研究对象，基于二元Logistic模型，从人、车、路、环境这4个方面，分别针对财产损失事故、伤人事故、死亡事故建立了模型。深入分析了道路物理隔离位置、路侧防护设施类型等因素对事故严重程度的影响，并利用Hosmer-Lemeshow检验和一致性检验对模型有效性进行验证。结果表明:①道路物理隔离的空间位置对事故严重程度有显著影响，仅布设中心隔离设施发生死亡事故的概率是同时布设中央和机非隔离的2.304倍。在有中心隔离设施的高等级道路中，增设机非隔离设施能有效降低事故发生的概率。②路侧防护设施类型为行道树、绿化带时，发生死亡事故的概率分别是金属护栏的1.982倍、1.648倍。与金属护栏相比，行道树更容易引发严重事故。③夜间无路灯照明发生死亡事故的概率是夜间有路灯照明的1.808倍，夜间无路灯照明是导致死亡事故的重要因素之一。④受过高等教育的驾驶人发生财产损失事故和伤人事故的概率较高，受过中等教育的驾驶员发生死亡事故的概率较高；受过中等教育驾驶员发生死亡事故的概率是高等教育驾驶员的2.049倍。研究深入分析了影响城市道路交通事故的显著因素及其对事故的影响，为事故严重程度的精细化分析提供了理论支持，为交通规划与管理部门提供了决策依据。 

An Analysis of Severity of Traffic Accidents on Urban Roadways Based on Binary Logistic Models

In order to accurately identify the factors affecting the severity of traffic accidents on urban roadways, 4 587 records from the traffic accident database of a city in China from Year 2018 to 2020 are used. A series of binary logistic models are developed for property damage accidents, personal-injury accidents, and fatal accidents based on the data regarding the following four aspects, including participants, vehicles, roads, and environment. The impacts of location of dividing strips and the types of roadside protection facilities on the severity of accidents are analyzed. Further, the Hosmer-Lemeshow tests and the consistency tests are used to check the soundness of the models. The results from the binary Logistic models show that: ① locations of dividing strips have a significant effect on the severity of the accidents. The probability of fatal accidents of placing central dividing strips only is 2.304 times higher than placing both central and motor/non-motor dividing strips. On high-grade roads with central dividing strips, adding motorized/non-motorized dividing strips can effectively reduce the probability of accidents. ② The probability of fatal accidents for the roadside protection facilities being street trees and green belts is 1.982 times and 1.648 times higher than the roadside protection facilities being metal guardrails, respectively. Street trees as the type of roadside protection facilities are likely to lead to more serious accidents than metal guardrails. ③The probability of fatal accidents at night "with no streetlight" is 1.808 times higher than that "with streetlights" at night. No streetlight at night is a significant risk factor leading to the fatal accidents. ④ Drivers with a high-level of education are more likely to have property-damage accidents and personal-injury accidents. Fatal accidents are more likely to occur among drivers with a medium-level of education, and their probability of resulting in the fatal accidents is 2.049 times higher than drivers with a high-level of education. In conclusion, this study presents an analysis of the factors and their impacts on traffic accidents at urban roadways. Furthermore, it can serve as a theoretical support for the refined analysis of accident severity on urban roadways, as well as a reference for traffic safety planning and management authorities.