时间序列的层次分段及相似性度量

Hierarchical segmentation and similarity measure of time series

时间序列的相似性度量是时间序列数据挖掘的研究基础,为数据挖掘任务的效率和准确度提供可靠的保障。提出一种时间序列的层次分段及相似性度量方法,方法首先识别时间序列中的极值点,依据极值点的特征对时间序列进行分层次分段,并以此为基础,通过定义新的距离公式来度量时间序列间的相似性。使用新提出的相似性度量方法对时间序列进行聚类计算,实验结果表明,该方法能够有效地度量时间序列间的相似性,聚类效果明显,具有较好的实用性和良好的应用前景。

Time series similarity measure is the basis of time series data mining, which assure the data mining jobs’efficiency and accuracy. This article proposes a hierarchical segmentation and similarity measure method of time series. The method spots extreme points in time series firstly, segments the time series hierarchically based on extreme points feature, and defines new distance formula to measure the time series’similarity. Applying new similarity measure method in time series cluster calculation, the experimental result shows the method can measure the time series similarity effectively and have evident cluster efficiency. The new similarity method is equipped with favorable practicability and well application prospect.