|
|||||
|
|
| 基于eEP的两阶段分类方法的研究 | |
| 引用本文： | 职为梅,周文刚,孙宜贵.基于eEP的两阶段分类方法的研究[J].商丘师范学院学报,2006,22(2):88-92. |
| 作者姓名： | 职为梅 周文刚 孙宜贵 |
| 作者单位： | 1. 郑州大学,计算机科学系,河南,郑州,450052 2. 北方工业大学,计算机科学系,北京,100041 3. 河南工业大学,计算机科学系,河南,郑州,450052 |
| 摘 要： | 传统的基于规则的分类算法多是采用顺序覆盖技术训练分类规则,这使得训练得到的模型覆盖大量的非目标类实例,分类时效果差.基于规则的两阶段分类算法,能够很好的去除模型覆盖的非目标类实例,分类时能取得比较好的结果.EP在分类大型数据库时能够提高分类效率,eEP(Essential Emerging Patterns)是一种特殊的EP,较EP能够减少分类噪音.本文中我们构造一个新颖的分类算法,基于eEP的两阶段分类方法(即EEPCTP),并使用UCI机器学习库中的10个数据集做实验,实验表明EEPCTP分类法取得了与一些经典的分类算法可比的效率和准确性. |
| 关 键 词： | 数据挖掘 分类 显露模式 两阶段分类 |
| 文章编号： | 1672-3600（2006）02-0088-05 |
| 收稿时间： | 2005-01-14 |
| 修稿时间： | 2005-03-18 |
Based on essential emerging patterns to classify in two phases |
|
| ZHI Wei-mei,ZHOU Wen-gang,SUN Yi-gui.Based on essential emerging patterns to classify in two phases[J].Journal of Shangqiu Teachers College,2006,22(2):88-92. | |
| Authors: | ZHI Wei-mei ZHOU Wen-gang SUN Yi-gui |
| Institution: | Department of Computer Science, Zhengzhou University, Zhengzhou 450052, China |
| Abstract: | Traditional algorithms for classification usually use technique of sequential covering to train model,but the methods can make the classifier cover many examples of the non-target class and affects the accuracy.Two Phase to Classify,which can remove most of the exemples of the non-target class that were covered by the classifier,can receive a good effect when classify rare class.EP is good for classification of large database,eEP is special EP which can reduce the noise of classification.In the paper we propose a novel approach,EEPCTP,which can be looked as a hybrid of eEP_based classifier and Two Phase classifier and use UCI Machine Learning Repository as experimental dataset.We prove that our method can achieve good effect. |
| Keywords: | data mining classification emerging pattern two phase classification |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ | |