交通事件持续时间分布拟合及其加速消散模型

Fitting Distribution of Freeway Incident Duration and Accelerated Failure Time Modeling

对某高速公路交通事件管理系统中记录的3年交通事件信息进行了数据统计处理,对交通事件持续时间的随机分布进行了多种拟合分析和分布检验,构建了基于对数逻辑斯特分布的交通事件持续时间加速消散模型.该预测模型可以使用不同的拟合分布并接受缺失数据,样本检验能很好地预测交通事件持续时间及其累积结束概率.

Distribution model choice,parameters＇ estimation and measures of goodness of fit were carried out,and the best distribution fitting for incident duration data was Log-Logistic distribution.Accelerated failure time（AFT） modeling was employed based on survival analysis and Log-Logistic distribution.Then stepwise by Akaike information criterion（AIC） was used to find the best group of factors for the prediction of freeway incident duration.The test result indicates that AFT model can represent actual incident duration and accept different distribution and missing data.