Abstract: Time series data is characterized as large in data size, high dimensionality and updates continuously. It is hard to manipulate for data analysis and mining in its original structure. Defining a more effective and efficient time series segmentation algorithm is of fundamental importance. This paper proposes a time series segmentation algorithm based on Series Importance Point (SIP), which can approximately represent time series by linear composed of SIP. This method adopts SIP as segmentation point in time series reflecting mostly character of time series. The dimensionality of time series is reduced, and the error of the whole is least.

Key words: time series, Series Importance Point(SIP), segmentation

摘要： 时间序列包含的数据量大、维数高、数据更新快，很难直接在原始时间序列上进行数据挖掘。该文提出一种基于序列重要点(SIP)的时间序列分割算法——PLR_SIP，用SIP组成的直线段近似描述时间序列。将SIP作为时间序列的分割点，反映时间序列的主要特征，降低时间序列的维数，使整体误差达到最小。

关键词: 时间序列, 序列重要点, 分割

CLC Number: 

• TP311