摘要
符号化时间序列聚类是聚类研究中的热点之一,其中关键问题是时间序列符号化相似度问题.本文针对传统的基于欧式距离度量存在的缺陷,以LCS度量为基础,提出了ELCS相似性度量,克服了LCS度量需要依赖线性函数选取的不足.在两类数据集上进行的实验表明,同其他常用度量的比较,该度量有着更好的聚类效果.
The symbolization time series cluster is one of hot spots in the clustering research.The key to question is similarity in signifying of time series.This paper puts forward the ELCS similarity measure based on the defects existed in the traditional Euclidean distance metric,and on foundation of the LCS measure.It can overcome the shortcomings of the LCS measure which rely on the selection of linear function.The experiment on two types of data sets shows that compared with other common measure,this measure has a better clustering effect.
出处
《微电子学与计算机》
CSCD
北大核心
2012年第11期74-77,共4页
Microelectronics & Computer
基金
教育部博士点基金项目(20100095110003)
中国矿业大学青年科技基金(2011QNB23)
