平均1-依赖决策树集成算法

Averaged One-Dependence Decision Trees Ensemble Algorithm

基于平均1-依赖贝叶斯分类器(AODE)算法的思想,提出了平均1-依赖决策树集成算法(AODT),该算法通过使用每个输入属性和类别属性共同建立集成学习中的个体决策树分类器.同时,我们从多任务学习的角度探讨了AODE和AODT算法的工作原理.通过在Weka平台上使用40个UCI数据集的实验结果表明,该算法可以显著提高决策树学习算法的分类性能,并且具有很好的抗噪声性能.

Averaged One-Dependence Estimators （AODE） ensemble naive Bayes classifiers by aggregating the predictions of a set of one-dependence estimators built for each attribute. Inspired by this, in this paper we propose a new method, namely Averaged One-Dependence Trees （AODT）, to ensemble decision tree teaming algorithms which enumerate each input attribute together with the class attribute to create different component one-dependence decision tree classifiers in the ensemble. We then give a multitask view of AODE and AODT to explain how they work. We conduct all the experiments on the Weka platform and use the 40 widely used UCI data sets. The experimental results verify the method＇s effectiveness, efficiency and robustness.