基于多变量决策树交通事件持续时间预测模型

A Traffic Incident Duration Time Predication Model Using Multivariable Decision Tree

基于粗集理论和决策树方法,建立了交通事件持续时间的多变量决策树预测模型。通过分析交通事件的属性特点,运用粗集理论中的属性约简方法,确定了交通事件的核心属性;运用等价关系相对泛化原理构造了多变量组合检验,并根据变量依赖度确定了最优变量组合;以多变量组合判据代替单变量判据建立了决策树模型,利用决策树高度和节点样本数对树的规模进行控制,优化了决策树结构。实例应用表明,该模型对交通事件持续时间的分类和预测能力较强,预测精度较高。

Using theory and method of rough set and decision tree,a multivariable decision tree model was developed for traffic incident duration time prediction.Through analyzing the incident attributes,the attribute reduction algorithm in rough set theory was used to get the core attributes of the incident.By using the generalization principle of equivalence relation,a multivariable combination test was formed.By comparing the dependence of different variable combinations,the optimal variable combination was determined.Then,multivariable combination criterion instead of single variable criterion was used to set up the decision tree,and through limiting tree height and number of tree leaves,the scale of tree was controlled,so,the tree＇s structure was optimized.The case study shows that this model has a good performance in classifying and forecasting traffic incident duration time,and it has good accuracy in duration time forecasting.