回归模型在交通事故与行车风险分析中的应用综述

A Comprehensive Review of the Application of Regression Models in Traffic Accident and Driving Risk Analysis

为提升道路交通安全,准确评估事故致因和行车风险因素至关重要。本综述旨在回顾和评估近年来应用于该领域的回归模型,使用VOSviewer软件对中外文献进行计量分析,以揭示该领域研究的趋势和热点。深入探讨一般线性模型、广义线性模型(GLM)、广义加性模型(GAM)和非线性模型在风险因素分析中的应用,尤其是详细讨论Logit回归模型、Probit回归模型和计数数据模型(如泊松模型和负二项模型)在实际数据集中的表现和适用性。此外,针对数据中多零值情况还探讨零膨胀模型(ZIP和ZINB)、多层膨胀模型、Lindley模型以及Tobit模型在处理样本选择偏误中的应用。特别强调不同模型对数据特性的处理能力,并基于模型特性和适用场景提供了科学的模型选择指导,旨在为研究人员在面对具体的交通安全问题时选择合适的回归分析模型提供参考。

Accurate evaluation of accident causes and driving risk factors is crucial for improving road traffic safety.This review aims to assess regression models applied in this field over recent years.Using VOSviewer for bibliometric analysis of both Chinese and international literature,the study reveals research trends and hotspots.It explores the application of general linear models,generalized linear models(GLM),generalized additive models(GAM),and nonlinear models in analyzing risk factors.Special attention is given to Logit regression,Probit regression,and count data models(such as Poisson and negative binomial models),evaluating their performance and applicability on real-world datasets.Additionally,the review addresses the issue of excess zeros in data,examining the application of zero-inflated models(ZIP and ZINB),hurdle models,Lindley models,and Tobit models in managing sample selection bias.Emphasizing the capability of different models to handle various data characteristics,the review provides scientific guidance on model selection based on their properties and applicable scenarios,aiming to assist researchers in selecting suitable regression analysis models for specific traffic safety issues.