一种基于重要点的时间序列分割方法

分段线性表示是时间序列降维的有效方法，其关键在于分割点的确定。在时间序列分段线性表示的基础上，提出一种新的基于重要点的时间序列分割方法。与一般方法比较相邻三点关系不同的是，将时间窗扩展为前一重要点、待考察点和一个指定时间窗组成的区间，再通过比较数据点前后模式变化来确定重要点。通过与其他7种分割方法进行实验比较，证明该方法适应能力强，不但分割结果总体质量高，在压缩率相同时具有更小的拟合误差，而且能够有效滤除噪声，发现时间序列的模式特征。

Method for time series segment based on important point

Piecewise linear representation is an effective method to reduce the dimension of time series.The crux of the meth-od is ascertaining the segmenting point.A new subsequences segment algorithm based on important point is proposed,refer-ence to the piecewise linear representation of time series.Different with the normal methods that compare the relation among the three neighboring points,the method’s check time window includes the former important point,the candidate point and the next specified time window.The important point is ascertained by comparing the change between the point front pattern and the back’s.Compared with other seven segmentation algorithms,the experiments show that the proposed algorithm has better performance,not only achieves the better overall quality on the segmentation results at the same compressibility,but al-so wipes off noise effectively and finds the pattern features of time series.