决策树的属性选取策略综述  

A Brief Survey of Decision Tree Attributes Selection Strategy

作　　者：韩松来[1] 张辉[1] 周华平[1] 

机构地区：[1]国防科学技术大学,长沙410073

出　　处：《微计算机应用》2007年第8期785-790,共6页Microcomputer Applications

摘　　要：决策树的属性选取标准在决策树算法中处于核心地位,针对数据集的不同特征人们提出了多种不同属性选取标准。该文首先回顾了几种常用的决策树属性选取标准,并把它们概括为属性间相互独立的选取策略和属性间相互关联的选取策略然后介绍了各个属性选取标准的多值偏向问题。Decision tree attributes selection criterion is the core of decision algorithm, and people have proposed many different criateria according to different data sets. Firstly, this paper reviews several frequently used attributes selection criteria, and generalizes them into two categories： attributes-lndependent strategies and attributes-dependent strategies. Finally, we give a short survey to the variety bias problem of attributes selection.

关 键 词：决策树 信息增益 GINI-INDEX X^2 统计 RELIEF CM 多值偏向 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]