基于粗糙集方法的决策树多值偏向理论分析  

The Theory Analysis of ID3 Algorithm Variety Bias Based Rough Set

作　　者：章晓[1] 何熊熊[1] 朱忠记 赵松[1] 

机构地区：[1]浙江工业大学信息工程学院,浙江杭州310023

出　　处：《杭州电子科技大学学报（自然科学版）》2014年第2期41-44,共4页Journal of Hangzhou Dianzi University：Natural Sciences

摘　　要：ID3决策树算法是构造决策树的重要算法之一,然而实验表明该算法在选择分裂属性时存在着多值偏向问题。以往的大多数学者都是基于实验分析多值偏向问题。该文针对这个问题,基于粗糙集理论及凹函数性质,引入函数重要度概念,从理论上分析多值偏向问题,并分析了属性多值对属性的重要度的影响。最后实验验证该理论分析方法的正确性与可行性。As one of the classical algorithms for building a decision tree,the ID3,has a problem of tending to choose the attribute which has more different values. This problem is also called multi-value bias. Most scholars used to research the multi-value bias based on the experiment. To address this problem,this paper,based on rough set theory and concave function,introduces a concept called attribute importance concept,presents a theoretical analysis of the multi-value bias problem. This paper further analyzes the influence of attribute＇s multi-value on the other attribute. Lastly,the result of experimental proves this theoretical method correct and feasible.

关 键 词：决策树算法 粗糙集 信息增益 多值偏向 属性重要度 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]