基于实例加权方法的概念漂移问题研究

被引:5
作者
胡学钢
潘春香
机构
[1] 合肥工业大学计算机与信息学院
基金
安徽省自然科学基金;
关键词
数据流; 概念漂移; 集成分类器; 分类;
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
数据流上的漂移概念发现已成为数据挖掘领域的研究热点之一。针对存在概念漂移的数据流分类问题,提出一种基于实例加权方法的数据流分类算法(EWAMDS),根据基分类器在训练实例上的分类结果调整该实例的权值,以增强漂移实例在新分类器中的影响,同时引入动态的权值修改因子以提高算法的适应性。实验结果表明,动态地调整实例的权值时算法的适应性更强;与weighted-bagging相比,EWAMDS的时间开销显著降低、分类正确率显著提高。
引用
收藏
页码:188 / 191
页数:4
相关论文
共 12 条
[1]  
Models and issues in data stream systems. Babcock B,Babu S,Datar M,et al. Proceedings of21st ACM Symposium on Prin-ciples of Database Systems . 2002
[2]  
A dynamic weighted majority:a new ensemble method for tracking concept drift. Kolrer J Z,Marcus. Proceedings of the3rd IEEE Conference on Data Mining . 2003
[3]  
Boosting classifiers for drifting con-cepts. Scholz M,Klinkenberg R. Intelligent Data Analysis . 2007
[4]  
Mining concept-drifting data streams using ensemble classifiers. Wang H,Fan Wei,Yu P,et al. Proceedings of9th ACM SIGKDD In-ternational Conference on Knowledge Discovery and Data Mining . 2003
[5]  
Using additive expert ensembles to cope with concept drift. Kolter J Z,Maloof M A. Proceedings of the22nd International Con-ference on Machine Learning . 2005
[6]  
Forest trees for on-line data. Gama J,Medas P,Rocha R. Pro-ceedings of the19th Annual ACM Symposium on Applied Com-puting . 2004
[7]  
Mining time-changing data streams. Hulten G,Spencer L,Domingos P. Proceedings of the7th ACM SIGKDD International Con-ference on Knowledge Discovery and Data Mining . 2001
[8]  
Dynamic classifier selec-tion for effective mining from noisy data streams. Zhu Xingquan,Wu Xingdong,Ying Yang. Proceedings of the4th IEEE International Conference on Data Mining . 2004
[9]  
Issues in data stream management. Golab L,Ozsu M T. Proceed-ings of the2003ACM SIGMOD International Conference on Man-agement of Data . 2003
[10]  
Fast and light boosting for adaptive mining of data streams. Chu F,Zaniolo C. . 2004