242.txt

This complete matlab for neural network

发信人: guitar (新宠酱牛肉), 信区: DataMining
标  题: Re: 请问为什么用小波变换对时间序列的相似性搜索
发信站: 南京大学小百合站 (Tue Dec 10 23:51:33 2002)

Wavelets are used for dimensionality reduction in order to support efficient
similarity search. Since time series are usually pretty long indexing
techniques such as R-trees, which are used to support fast similarity search,
break down due to the notorious overlapping problem in high-dimensional
spaces. To alleviate this problem, we can use 
wavelets to obtain a transform of the original time series, and then keep only 
a few high level coefficients, which effectively reduce the dimensionality. 
After that, we can index the reduced data (the coefficients) with traditional techniques. 

【 在 xiaojing2002 的大作中提到: 】
: 其实，用小波变换对时间序列进行处理，在数据量方面与原始数据是等价的（即使小..
: 换可以除去噪声）。为什么不直接用欧氏距离公式进行判断？
: 而且好像国内对时间序列在数据挖掘方面的研究不是很多，基本都是统计系的再搞这..
: 的研究，请问是否听说比较有名的这方面的专家，谢谢回答

--
有时候　有时候　
我会相信一切有尽头　
相聚离开　都有时候　
没有甚麽会永垂不朽

※ 来源:．南京大学小百合站 http://bbs.nju.edu.cn [FROM: 64.231.249.107]

※ 修改:．guitar 於 Dec 10 23:53:23 2002 修改本文．[FROM: 64.231.249.107] 
※ 修改:．guitar 於 Dec 10 23:53:50 2002 修改本文．[FROM: 64.231.249.107] 
※ 修改:．guitar 於 Dec 10 23:55:44 2002 修改本文．[FROM: 64.231.249.107] 
※ 修改:．guitar 於 Dec 10 23:56:34 2002 修改本文．[FROM: 64.231.249.107]