变换域时间序列相似性搜索

Time series similarity search based on transformation

在时间序列的研究中,经常需要计算二个序列的相似程度。由于序列变化的多样性和复杂性,结果通常不能很好地满足要求。采用变换法则对时间序列进行从时域到频域的转换,再将转换后的数据按照一定的规则变换成字符序列;利用求最长公共子序列的方法计算二个序列的匹配度,实现时间序列的相似性搜索。

In the time-series searching,similarity of the two series is always involved.Because of the diversities and complication of the series,the results cannot meet the needs well.This paper proposes a method by translating the series from time domain to frequency domain,then to a string sequence.By calculating the longest common subsequence ,finally,similarity searching in time-series data sets is realized.