中文摘要：

为分析显著影响追尾碰撞事故的因素,分别采用负二项（NB）回归模型和零堆积负二项（ZINB）回归模型建立追尾碰撞事故起数预测模型。通过模型的拟合优度检验和准确性检验,对比2种模型的优劣,采用弹性分析确定自变量对因变量的影响程度。以2006—2008年某高速公路发生的追尾碰撞事故进行实证性研究。结果表明：ZINB回归模型比NB回归模型具有更好的拟合效果和预测准确性;在ZINB回归模型中,曲度、曲线比例、车道宽度与追尾碰撞事故起数显著正相关,曲率变化率、相邻路段坡差和弯坡组合与追尾碰撞事故起数显著负相关;车道宽度、曲率变化率和曲度对追尾碰撞事故的发生影响尤为明显。

英文摘要：

In order to analyze factors influencing significantly rear-end collision accidents, two traffic accident frequency prediction models were established based on NB regression model and ZINB regression model respectively. According to results of goodness-of-fit and prediction accuracy, a comparative analysis was carried out between two prediction models. The elastic analysis method was used to determine influ- ence degree of the independent variables on the dependent variable. Rear-end collision accidents occurred on a certain expressway from 2006 to 2008 were taken as an example. The results show that rear-end colli- sion accident frequency prediction model based on ZINB model is better than that based on NB model, that in ZINB model, their is positive correlation of curvature, the proportion of curve and width of lanes with rear-end collision accident frequency, and there is a significant negative correlation of curvature variation, gradient on adjacent road and he combination of bending slope have with rear-end collision accident frequency, and that width of lanes, curvature variation and curvature have a significant effect on occurrence of rear-end collision accident.