Abstract: The precision of the similarity measurement standard in index space is lower, in the existing available similar time-subsequence searching algorithms, the time for I/O of data is longer in the post processing. A non-linear dimensionality reduction technique is proposed, and a new algorithm of similar subsequence search in time series is presented. The precision of the similarity measurement standard in this algorithm is enhanced, and the rate of false match is decreased effectively. Experimental results demonstrate that the algorithm can decrease the amount of data by repeat-estimation effectively and improve the efficiency of the similarity search.

Key words: time series, dimensionality reduction, subsequence, similarity search

摘要： 索引空间相似性度量标准的精确度较低，在现有的相似时间子序列搜索算法中，后期处理过程中内外存之间数据的I/O时间较长。针对该问题，引用一种非线性维数约简技术，提出改进的相似时间子序列快速搜索算法，提高索引空间相似性度量标准的精确度，降低错误匹配的发生率。实验结果表明，该算法可有效降低重复估算的数据量，提高相似性搜索的效率。

关键词: 时间序列, 维数约简, 子序列, 相似性搜索

CLC Number: 

• TP311.52