摘要
在时间序列数据挖掘中,传统的时间序列相似性度量算法没有考虑反映时间序列结构的关键点特征。为了解决该问题,文章提出了基于波动特征的时间序列相似性度量算法,并通过聚类进行了效果分析。研究中首先利用小波分析方法提取时间序列整体变化趋势,然后给出了针对小波分析得到的序列进行波动点识别的方法,构造出包含时间序列重要波动信息的波动点序列。最后提出了非等长时间序列的相似性度量方法计算波动点序列间的距离。实验结果表明,该时间序列度量方法能更好地反映时间序列的趋势特征。
In time series data mining,traditional time series similarity measurement algorithms do not consider the features of key points reflecting the structure of time series. In order to solve this problem,this paper proposes a time series similarity measurement algorithm based on fluctuation features,and analyzes the effect by clustering. Firstly,the wavelet analysis method is used to extract the global variation trend of time series. Then,the method of identifying the fluctuation points for the wavelet series is given,and the fluctuation point series with important fluctuation information of the time series is constructed. Finally,a similarity measurement method for non-equal time series is proposed to calculate the distance between fluctuation points. The test results show that the time series measurement method can better reflect the trend characteristics of time series.
作者
陈海兰
高学东
Chen Hailan;Gao Xuedong(Donlinks School of Economics and Management,University of Science and Technology Beijing,Beijing 100083,China)
出处
《统计与决策》
CSSCI
北大核心
2019年第11期17-22,共6页
Statistics & Decision
基金
国家自然科学基金资助项目(71272161)
关键词
时间序列
相似性度量
聚类
波动特征
time series
similarity measure
clustering
fluctuation characteristics
