灾害天气下高速公路风险评估的Logit模型改进

Improved Logit Model of Risk Assessment for Freeway in Disaster Weather

Logit模型使用时要求自变量数量少、相互独立.灾害天气下高速公路风险评估的自变量种类和数量较多,变量之间存在相关性.因此,利用主成分分析将多个相关的自变量转化为少数、相互独立且包含绝大部分原始信息的新自变量,避免自变量共线性对Logit模型的影响.主成分分析表明,风险分析自变量相关性较强,不能直接使用Logit建模;主成分依次为:天气、交通流、交通标志,灾害天气的等级和持续时间是最主要变量,二者负相关,交通量与车型比例正相关,标志的清晰度和可视性正相关.Logit模型表明,天气、交通流与风险等级正相关,标志标线与风险等级负相关.实测数据证明改进模型比Logit模型评估精度更高;风险等级越高时,改进模型评估精度越高.

When the logit model is used,the number of independent variables should be limited and independent.In disaster weather,there are many kinds and quantities of independent variables in highway risk assessment,and there are correlations among variables.Therefore,using principal component analysis,a number of related independent variables were transformed into a few new mutual independent variables containing most of the original information,to avoid the influence of collinearity of independent variables on the logit model.Principal component analysis shows that the independent variable of risk analysis is relatively strong and cannot be modeled directly by logit model.The main components include the weather,traffic flow and traffic marking.The level and duration of disaster weather are the most important variables,which present the negative correlation.The traffic volume is positively related to vehicle type,and the clarity and the visibility of the marking present the positive correlation.The logit model shows that the weather and traffic flow are positively related to the risk level,and the marking is negatively related to the risk level.The measured data show that the improved model is more accurate than the logit model.The higher the risk level is,the higher the prediction accuracy of the improved model is.