Abstract:
This paper proposes a subsequences partition algorithm for different width time window, focusing on the shortage of that by the same width time window. Aimed at dealing with the limitation of the approach, a new approach that divides the time series with different width time window is put forward. Due to the length of subsequences obtained by the new approach is different, the similarity between these subsequences can only be measured by the algorithm of DTW, while its calculation speed is slow. So a new distance measurement algorithm, nonuniform time series (NTS) is put forward. The approach and the algorithm are tested. Experimental result shows the superiority of the new approach and new algorithm.
Key words:
time window,
time series,
clustering,
data mining
摘要: 针对相同时间窗对时间序列进行子序列划分的缺点,提出一种异时间窗的子序列划分方法。为解决划分得到的子序列长度不同,而使用动态时间弯曲算法进行子序列相似性度量的计算速度慢的问题,给出一种不规则时间序列距离度量算法。对异时间窗的子序列划分方法和不规则时间序列距离度量算法进行了实验,结果证明了二者的优越性。
关键词:
时间窗,
时间序列,
聚类,
数据挖掘
CLC Number:
GUO Hong-wei; GAO Xue-dong; WANG Hong. Time Series Clustering Based on Different Width Time Window Partition[J]. Computer Engineering, 2007, 33(21): 3-5.
国宏伟;高学东;王 宏. 基于异时间窗划分的时间序列聚类[J]. 计算机工程, 2007, 33(21): 3-5.