中国安全科学学报 ›› 2020, Vol. 30 ›› Issue (8): 129-136.doi: 10.16265/j.cnki.issn1003-3033.2020.08.019

• 公共安全 • 上一篇    下一篇

基于混合聚类的农村公路单车事故影响因素分析

杨慧敏, 石琴 教授, 陈一锴 副教授, 骆仁佳   

  1. 合肥工业大学 汽车与交通工程学院,安徽 合肥 230009
  • 收稿日期:2020-05-22 修回日期:2020-07-13 出版日期:2020-08-28 发布日期:2021-07-15
  • 作者简介:杨慧敏 (1994—),女,安徽太和人,硕士研究生,研究方向为道路交通安全。E-mail: huiminyang_2019@163.com。
  • 基金资助:
    国家自然科学基金资助(71871078)。

Influencing factors for single vehicle accidents on rural highways based on hybrid clustering approach

YANG Huimin, SHI Qin, CHEN Yikai, LUO Renjia   

  1. School of Automotive and Transportation Engineering, Hefei University of Technology, Hefei Anhui 230009, China
  • Received:2020-05-22 Revised:2020-07-13 Online:2020-08-28 Published:2021-07-15

摘要: 为探究影响安徽省农村公路单车事故严重度的主要因素,利用因子分析法,将自变量转化为相互独立的公共因子,依据因子得分,利用K均值算法聚类事故数据;采用二元Logistic回归模型对各类别数据建立事故严重度模型。结果表明:相对于潜在类别分析,基于混合聚类结果构建的Logistic回归模型拟合优度、预测精度更优;性别、年龄、是否超速等仅在某一类别中显著;道路线形、地形等在多个类别中显著,但对于事故严重度的影响方向不同。

关键词: 农村公路, 单车事故, 潜在类别分析, 因子分析, K均值聚类, 二元Logistic回归

Abstract: In order to explore key factors that affect severity of single vehicle crashes on rural highways in Anhui Province, factor analysis was employed to transform independent variables into independent common factors. Then, K-means algorithm was used to cluster crash data according to factor scores. Finally, a binary Logistic regression model for accident severity was developed for each cluster. The results indicate that compared with latent class analysis, Logistic regression model, based on hybrid clustering results, has better goodness-of-fit and higher prediction accuracy. Factors such as gender, age and overspeed are only significant in a certain cluster while road alignment and terrain are significant in many, but exert different influence directions on crash severity.

Key words: rural highway, single vehicle accident, latent class analysis, factor analysis, K-means clustering, binary Logistic regression

中图分类号: