| 基于信息增益总体竞争的决策树分裂属性选择算法 |
| |
| 引用本文： | 郭世仁,简友光.基于信息增益总体竞争的决策树分裂属性选择算法[J].仲恺农业技术学院学报,2008,21(1):42-45. |
| |
| 作者姓名： | 郭世仁 简友光 |
| |
| 作者单位： | 1. 仲恺农业技术学院,信息学院,广东,广州,510225
2. 仲恺农业技术学院,继续教育学院,广东,广州,510225 |
| |
| 基金项目： | 仲恺农业技术学院校级教学研究基金 |
| |
| 摘 要： | 提出了一种能在一定程度上避免决策树陷入局部最优的分裂属性选择算法：总体竞争.该方法考虑了候选属性在总体学习样本上的分类能力,以此通过各候选属性的相互竞争,确定决策树增长过程中的分裂属性.以Iterative Dichotomizer 3（ID3）算法为对照的实验数据表明,该方法能以较小的代价获得较高的决策树准确率.
|
| 关 键 词： | 分类 决策树 信息增益 总体竞争 ID3 |
| 文章编号： | 1006-0774(2008)01-0042-04 |
| 修稿时间： | 2007年10月10 |
An algorithm for selecting split attribution in decision tree: ensemble vote based on information gain |
| |
| GUO Shi-ren,JIAN You-guang.An algorithm for selecting split attribution in decision tree: ensemble vote based on information gain[J].Journal of Zhongkai Agrotechnical College,2008,21(1):42-45. |
| |
| Authors: | GUO Shi-ren JIAN You-guang |
| |
| Institution: | GUO Shi-ren, JIAN You-guang (1. College of Information, Zhongkai University of Agriculture and Technology, Guangzhou 510225, China; 2.College of Continuing Education, Zhongkai University of Agriculture and Technology, Guangzhou 510225, China) |
| |
| Abstract: | An algorithm for selecting split attribution in decision tree with ensemble vote was proposed to avoid local optimization to a certain extent.The classification capability of these candidate attributions on ensemble sample was calculated,and the candidates were voted to select a split attribution to build decision tree.Contrast to Iterative Dichotomizer 3(ID3),the results of experiments showed that the method improved the predictive accuracy at a low cost. |
| |
| Keywords: | classification decision tree information gain ensemble vote ID3 |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
