决策树ID3算法的分析与优化  

ID3 algorithm for decision tree analysis and optimization

作　　者：黄宇达[1,2] 范太华[1] 

机构地区：[1]西南科技大学计算机科学与技术学院,四川绵阳621010 [2]周口职业技术学院信息工程系,河南周口466000

出　　处：《计算机工程与设计》2012年第8期3089-3093,共5页Computer Engineering and Design

基　　金：河南省教育厅自然科学研究计划基金项目(2008B520047)

摘　　要：对ID3算法的基本原理及其主要不足以及现有几种改进算法的优缺点进行了简要分析,针对ID3算法的主要不足即倾向于多值属性的选取,利用粗糙集理论和数学相关知识点对其进行了一定程度的改进。理论分析和实验结果表明,改进后的算法在一定程度上不仅较好地解决了ID3算法的多值偏向问题而且大大简化了算法的计算过程,明显提高了算法分类准确度和执行效率。First, ID3 algorithm＇s basic principles and major shortcomings, and advantages and disadvantages of several existing improved algorithms are simply analyzed by this paper. Then for ID3 algorithm the main drawback that tends to select the attribute which has more values, which has been significantly improved by using the rough set theory and mathematical knowledge points. Theoretical analysis and experimental results show that the improved algorithm, to a certain extent, not only can well solve the multi--valued bias problem of ID3 algorithm and greatly simplify the computational process, obviously improve the algorithm＇s classification accuracy and implementation efficiency.

关 键 词：决策树 ID3算法 信息熵 粗糙集 客观属性重要度 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]