一种基于趋势的时间序列相似性模型

Similarity model based on trend of time series

针对传统基于欧氏距离的时间序列相似性模型,搜索相似时间序列子段效率低、形态不完整的问题,本文提出了一种基于趋势的时间序列相似性模型。该模型以时间序列分段线性表示和符号化为基础,通过计算符号串的编辑距离得出时间序列在形态上的差异性。同时该模型把时间序列长度和时间序列变化幅度单独考虑,扩大了该模型的使用范围。实验表明,本文提出的基于趋势的时间序列相似性模型,在相似性匹配方面,搜索效率高于传统基于欧氏距离时间序列相似性模型;同时,该模型侧重于形态方面的相似性,所以对时间轴方向的偏移和白噪声不敏感,适用于不等长时间序列的相似性匹配。

In view of the problem that the time series similarity model based on Euclidean distance is less efficient and incomplete in morphology ,the paper presents a similar model based on trend for time series .The model is based on piecewise linear representation and symbolic representation of time series ,and the difference of time series is obtained by calculating the edit distance of the symbol string .At the same time ,the model considers the time series length and the time series variation separately ,expanding the use of the model .The experimental results show that the similarity model of time series based on trend is suitable for the similarity matching ,and the efficiency is higher than traditional model which is based on Euclidean distance .Since the model emphasized on the similarity of morphology ,it’s not sensitive to the deviation of time axis and white noise ,and it’s suitable for the similarity matching of time series with different length .