基于关联矩阵的决策树分类算法  

A New Decision Tree Algorithm Based on Association Matrix

作　　者：方立 

机构地区：[1]兰州市经济管理干部学校,甘肃兰州730083

出　　处：《长春大学学报》2013年第4期426-429,共4页Journal of Changchun University

摘　　要：ID3算法是决策树学习归纳和数据挖掘中的核心方法。针对ID3算法存在的多值偏向问题,该文提出了一种新的方法对ID3算法加以改进。首先建立属性的关联矩阵,然后通过计算属性的类方差选择分裂属性,结合实例说明了改进算法的基本思想。实验结果表明,改进后的算法能够构造更合理的决策树并能在一定程度上克服多值偏向。ID3 algorithm is a key method in induction and data mining. To solve the muhi-value bios problem in ID3 algorithm, this paper presents a new method, which establishes the association matrix with attribution, selects the splitting attribute through calculating class variance and explains the basic thinking with examples. Experimental results prove that the improved algorithm can construct a better decision tree and avoid the multi-value bios to a certain degree.

关 键 词：决策树 ID3 多值偏向 关联矩阵 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]