基于变换的时间序列相似性分析

为了解决时间序列相似性比较问题,采用从时间序列的直观特征分析入手进行定义的方法,定义了具体的基于变换的时间序列的相似性,并分析了良好的时间序列变换函数所应具备的性质,讨论了一些有代表性的基于变换的时间序列相似性的定义和分析方法,对这些方法的基本思想加以提炼和总结,并讨论了这些方法的优、缺点,为基于变换的现代时间序列分析方法研究提供了较为完整的成果概览。同时,提出了借助变换函数来对时间序列的相似性进行定义的方法,为进一步做好时间序列相似性的比较工作提供了具体方向和理论依据。

Approaches for time series similarity analysis based on transformation

To solve the problem of time series similarity comparison,the paper uses the method of studying time series with intuitive features and gives the definition of time series similarity based on the transformation from the shape features of time series.Then the paper gives the analysis of what natures a good time series transformation function should have.A number of representative time series similarity analysis methods are discussed based on the transformation.The basic ideas of these methods are analysed and the advantages and disadvantages of these methods summarized.A more comprehensive overview of modem time series analysis method is given based on the transformation.At the same time,the definition using transformation functions is given and research directions and theoretical basis are provided for the time series similarity comparison work.