基于多元回归分析的事件持续时间预测

Prediction of Incident Duration Based on Multiple Regression Analysis

事件持续时间预测是实现先进的事件管理系统的一个重要基础。本文通过多元回归分析法对高速公路事件持续时间预测问题进行研究。首先通过方差分析(ANOVA)确定对事件持续时间具有显著影响的因素。在此基础上,通过逐步回归分析确定用于高速公路事件持续时间预测的最佳变量组合并建立多元线性回归模型。经检验,预测值与实际值的相关系数为0.8573,预测结果基本能够反映真实的事件持续时间情况。

The prediction of incident duration is an impoaant foundation of expressway incident management system. In this paper, the problem of expressway incident duration prediction is studied. First, the marked influence factors of incident duration are determined by analysis of variance （ANOVA） .Then, stepwise regression analysis is used to select a best group of factors for the prediction of expressway incident duration, and the multiple linear regression model is established. Finally, the prediction effect of the regression model is tested.The test result indicates that the correlation of the prediction values and factual values is 0.8573, so the test result can generally represent actual incident duration.