基于判别分析的高速公路交通安全实时评价指标  

Discriminant analysis based method to develop real-time crash indicator for evaluating freeway safety

作　　者：徐铖铖[1] 刘攀[1] 王炜[1] 蒋璇[1] 

机构地区：[1]东南大学交通学院,南京210096

出　　处：《东南大学学报（自然科学版）》2012年第3期555-559,共5页Journal of Southeast University：Natural Science Edition

基　　金：国家自然科学基金资助项目(50908050);国家重点基础研究发展计划(973计划)资助项目(2012CB725402);江苏省研究生科研创新计划资助项目(CXZZ_0164);教育部博士生学术新人奖资助项目

摘　　要：为了实时评价交通流变化对高速公路交通事故风险的影响,利用高精度交通流数据建立了高速公路交通安全实时评价指标.提取了美国加州I-880 N高速公路上采集间隔为30 s的实时交通流数据和事故数据,采用Fisher判别分析方法建立交通流参数的线性组合,以判别危险交通流状态.该线性组合被定义为高速公路交通安全实时评价指标,当该指标小于0时,代表危险交通流状态,有发生交通事故的风险;当该指标大于0时,代表正常交通流状态,没有发生交通事故的风险.采用条件logistic回归模型进一步研究了该指标与交通事故风险之间的定量关系.研究结果表明,该指标预测事故的精度为65.7%.另外,该指标每减小1个单位,交通事故风险将提高1.8倍.To evaluate the impacts of changing traffic flow conditions on the risks of crash on free- ways, a quantitative indicator based on real-time traffic flow data is developed to predict the occur- rences of crash on freeways. Crash data and real-time traffic data with an interval of 30 s are extrac- ted from a segment of the 1-880 N freeway in the state of California, United States. A linear combi- nation of traffic flow parameters which is defined as real-time crash indicator is developed by Fisher discriminant analysis to identify dangerous traffic conditions. Negative real-time crash indicators represent dangerous traffic conditions potentially leading to crash occurrences while positive crash in- dicators represent normal traffic conditions which theoretically will not lead to crash occurrences. The conditional logistic regression model is applied to quantify the impacts of real-time crash indica- tor on the crash risk. The results show that the accuracy of using real-time crash indicator to predict crash occurrences on freeways is 65.7%. Moreover, one unit decrease in real-time crash indicator may increase the crash risk by 1.8 times.

关 键 词：高速公路交通安全 FISHER判别分析 实时事故风险 交通事故 条件logistic回归模型 

分 类 号：U491.31[交通运输工程—交通运输规划与管理]