After introducing many methods, this paper illuminates two pre-pruning means: PDTBS and PDTBP designed by author, other experiment based on UCI data show two algorithms can prune decision-tree to large extent on the condition of that accuracy diminish hardly.
本文介绍了许多事前、事后修剪算法,并阐述了本人提出的两种事前修剪算法:基于结点支持度的事前修剪算法PDTBS和基于结点纯度的事前修剪算法PDTBP。 在另一个基于UCI数据的实验中实现了提到的几种修剪算法以及PDTBS和PDTBP,结果表明:后两者在对树分类精度影响极小的条件下,大幅度地修剪了决策树。
参考来源 - 决策树的结点属性选择和修剪方法研究·2,447,543篇论文数据,部分数据来源于NoteExpress
应用推荐