摘要
首先对决策树ID3算法基本原理及主要不足进行了简要分析,然后针对其主要不足即分裂属性选取过程中的多值偏向问题,通过引入一种修正函数对其加以改进,同时又提出了一种独立性假设。理论分析和实验结果表明：改进算法在一定程度上不仅较好地弥补了多值偏向的最大不足,而且还大大简化了算法计算过程,在提高分类准确度的同时也明显加快了决策树构建速度。
First,ID3 algorithm＇s basic principles and major shortcomings have been analyzed simply,and then for the main shortcoming of ID3 algorithm that tends to select a attribute which has many values in the course of selecting split-properties,and then the ID3 algorithm has been improved by introducing a correction function and Proposing a hypothesis of independence.Theoretical analysis and experimen tal results show that the improved algorithm,to some extent,not only better compensate for the lack of multi-valued bias of the largest,but also greatly simplifies the algorithm process,improve the classification accuracy significantly and accelerate the speed of decision tree construction.
出处
《电脑知识与技术》
2012年第1期96-98,共3页
Computer Knowledge and Technology
基金
河南省教育厅自然科学研究计划项目（2011A520052）
关键词
决策树
ID3算法
修正函数
独立性假设
加权独立信息增益
decision tree
ID3 algorithm
correction function
the assumption of independence
weighted independent information gain
