一种基于权重的时间序列相似性度量  

A WEIGHT-BASED SIMILARITY METRICS FOR TIME SERIES

作　　者：江诗锋[1] 何振峰[1] 

机构地区：[1]福州大学数学与计算机科学学院,福建福州350108

出　　处：《计算机应用与软件》2010年第9期116-118,共3页Computer Applications and Software

基　　金：福建省青年创新基金(2006F3075)

摘　　要：依据时间序列的形态特征,为基于欧氏距离的相似性度量加入奖惩因子,使其能反映序列形态的相似性。同时根据相关的背景知识给时间序列不同的维设定不同的权重,并给出一种自动求权重集合的算法。该算法成功应用于福泉高速行车数据以及仿真数据的相似性度量。In this paper we added the rewards and punishment factor into the Euclidean similarity metrics according to the morphological characteristics of time series to enable it being able to represent the similarity of the series morphology. Besides, we set different weight to dif- ferent dimensions of time series according to corresponding background knowledge, and proposed an algorithm for automatic weight set search. The algorithm has been applied to the similarity metrics of Vehicle Traffic Data of the Fuquan Expressway and its simulation successfully.

关 键 词：时间序列 权重 形态 相似性 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]