A Modified Algorithm for Missing Values in Data Stream Decision Tree Classification

ISCAS OpenIR

	A Modified Algorithm for Missing Values in Data Stream Decision Tree Classification
	Hou, Xu-shan; Lv, Pin; Wang, Hao
	2014
会议名称	International Conference on Artificial Intelligence and Software Engineering (AISE)
页码	307-313
会议日期	JAN 11-12, 2014
会议地点	Phuket, THAILAND
收录类别	CPCI
出版地	DESTECH PUBLICATIONS, INC
ISBN	978-1-60595-150-8
部门归属	[Hou, Xu-shan; Lv, Pin; Wang, Hao] Chinese Acad Sci, Inst Software, Sci & Technol Integrated Informat Syst Lab, Beijing 100190, Peoples R China.
摘要	Missing values in data stream will affect the accuracy of classification. It is significant to research how to handle missing values in data stream decision tree classification. Considering the time performance of an existing algorithm is not good enough, a more efficient algorithm is proposed in this paper. The modified algorithm makes improvements in two aspects: on the one hand, according to the standard deviation of attribute value, we select different process methods for missing values, which could reduce the time complexity; on the other hand, we optimize the update mechanism, which could reduce the update time. The experiment results show that the run-time of our algorithm is reduced by 20%-70%, while the accuracy is the same as the existing algorithm.; Missing values in data stream will affect the accuracy of classification. It is significant to research how to handle missing values in data stream decision tree classification. Considering the time performance of an existing algorithm is not good enough, a more efficient algorithm is proposed in this paper. The modified algorithm makes improvements in two aspects: on the one hand, according to the standard deviation of attribute value, we select different process methods for missing values, which could reduce the time complexity; on the other hand, we optimize the update mechanism, which could reduce the update time. The experiment results show that the run-time of our algorithm is reduced by 20%-70%, while the accuracy is the same as the existing algorithm.
关键词	Data Stream Classification Decision Tree Missing Values
语种	英语
内容类型	会议论文
URI标识	http://ir.iscas.ac.cn/handle/311060/16520
专题	中国科学院软件研究所
推荐引用方式 GB/T 7714	Hou, Xu-shan,Lv, Pin,Wang, Hao. A Modified Algorithm for Missing Values in Data Stream Decision Tree Classification[C]. DESTECH PUBLICATIONS, INC,2014:307-313.