首页 | 本学科首页   官方微博 | 高级检索  
     

面向数据挖掘的时间序列符号化方法研究
引用本文:李斌,谭立湘,章劲松,庄镇泉.面向数据挖掘的时间序列符号化方法研究[J].电路与系统学报,2000,5(2):9-14.
作者姓名:李斌  谭立湘  章劲松  庄镇泉
作者单位:中国科学技术大学电子科学与技术系,合肥,230026
基金项目:973国家重点基础研究发展规划项目!(G1998030413)支持。
摘    要:针对时间序列的数据挖掘首先需要将时间序列(Time Series)数据转换为离散的符号序列(Symbol Sequences),本文提出了一个简单高效的时间我符号化方法,该方法的特点:一是利用线性化分段表示法所独有的形态分割与表达能力实现了时间序列的分段与表示:二是利用神经网络模糊聚类算法实现了时间序列的在线聚类。提出以矢量间开矿相似性度量作为聚类依据。并利用该方法实现了对金融领域时间序列数据的符

关 键 词:时间序列  符号化  数据挖掘  线性化分段  神经网络
文章编号:1007-0249(2000)02-0009-06

The Study of the Data Mining Oriented Method for the Symbolization of Time Series
LI Bin,TAN Li-xiang,ZHANG Jin-song,ZHUANG Zhen-quan.The Study of the Data Mining Oriented Method for the Symbolization of Time Series[J].Journal of Circuits and Systems,2000,5(2):9-14.
Authors:LI Bin  TAN Li-xiang  ZHANG Jin-song  ZHUANG Zhen-quan
Abstract:In the process of data mining on time series, continuous time series needs to be transformed into discrete symbolic sequences. This paper proposes a simple but very efficient method for the symbolization of time series. This method has two features: First, it segments the time series by an approach called linear segmentation, which is well known for its shape expression and segmentation. Secondly it implements the online clustering analysis on the segmented time series with a fuzzy clustering algorithm using neural network. The measure of shape similarity is also proposed in this paper. Effectiveness of this method has been verified by the symbolization of sample financial time series.
Keywords:Time Series  Symbolization  Data Mining  Linear Segmentation  Neural Netwo
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号