决策树算法中的连续属性处理方法  

The Study of the Discretization Method in Decision Tree

作　　者：许俊[1] 

机构地区：[1]南京工程学院计算机工程系,江苏南京210013

出　　处：《河北理工学院学报》2007年第2期71-74,80,共5页Journal of Hebei Institute of Technology

摘　　要：决策树是分类数据挖掘的重要方法。C4．5算法延用了经典ID3算法的基本策略,增加了处理连续属性的方法。在C4．5算法的基础上,讨论了新的基于属性变换的连续属性处理方法。该方法基于统计概率信息,依据概率属性的最佳分裂对应分裂连续属性,增加了决策树的分类精度。Decision tree is one of the important Categorising Data Mining methods. Algorithm C4. 5 makes use of the basic strategy of classical algorithm ID3, and makes the dicretization method joined as well. Based on algorithm C4. 5, this paper has put forward a new discretization method based on attribute transforming. According to statistic information, this method contact splitting numeric attribute with the best splitting probability attribute, which can enhance the categorising accuracy.

关 键 词：决策树 CA.5算法 连续属性 概率属性 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]