基于决策树的高速公路事件持续时间预测

利用决策树方法对高速公路事件持续时间预测问题进行研究。首先在借鉴各国研究经验的基础上，根据所研究事件数据集中的事件持续时间数据的分布特征确定构造基于决策树的预测方法；然后用整理得到的660组事件数据，通过对各类事件的显著性分析，建立高速公路事件持续时间预测决策树，并用同一数据集中未用于决策树构造的170组数据对决策树的预测效果进行检验。检验结果表明：所开发决策树的预测值与实际值的相关系数为0．8423，预测结果基本能够反映真实的事件持续时间情况。

Prediction of freeway incident duration based on decision tree

In the paper, the problem of freeway incident duration prediction is studied by decision tree method. First, based on the experiences of researches at home and abroad and distribution condition of incident duration in the data set used by the authors, the prediction method to develop decision trees is determined. Then, 660 teams of actual freeway incident data are used to develop freeway incident duration prediction method on decision trees by significance analysis on all kinds of incidents, and 170 teams of incident data in the same data collection, which are different from modelling data are used to test the prediction effect of the decision trees. The test result indicates that the correlation coefficient of prediction values and actual values is 0. 842 3 and it shows that the prediction result of the developed prediction method on decision trees can basically represent actual incident duration.