摘要
针对基于滑动窗口的互相似流群在线发现这一新的流分析问题,提出一种基于Lp-norm的相似性度量L’p-norm,证明了L’p-norm度量具有对称性、增量性和限制阈值下的传递性特点.基于L’p-norm,提出一种有效降低两两计算次数,增量计算的高效互相似流群发现算法ESSG.实验表明,ESSG的运行效率适用于在线分析.
To solve the new research problem of online discovery of mutual similar stream group in the area of data stream analysis, the paper put forwards a new similarity measure LI p-norm based on Lp-norm metrics, and shows its characteristic of symmetric, incremental computation and transitive with the limitation of threshold. Based on L' p-norm, an effective algorithm (ESSG) for mutual similar stream group discovery is proposed to reduce the computation of pair of streams and to achieve the incremental computation. With the extensive experiments, ESSG shows better performance applicable for online anaiysis.
出处
《小型微型计算机系统》
CSCD
北大核心
2007年第7期1245-1248,共4页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(60573090)资助
辽宁省自然科学基金项目资助
关键词
数据流
滑动窗口
相似性
data stream
sliding window
similarity