数字趋势序列及其全序列匹配算法研究

Research on Number Trend Sequences and Whole Sequence Matching Algorithm

针对基于点距离的时序数据分析和传统趋势序列分析的缺点,提出了数字趋势序列、序列的Lp距离、序列分段向量等概念,证明了包括“序列分段均值定理”在内的3个重要定理,设计了专门用于数字趋势序列的“基于序列分段向量(SSV)的全序列匹配算法”。算法使用片段斜率所对应的弧度值来度量片段的趋势,同时用趋势的保持时间来对趋势值进行加权,实现了数字趋势序列之间快速的全序列相似性搜索。

To overcome the demerits of point-distance-based temporal data analysis and traditional trend sequence analysis, the concepts of number trend sequence, Lp distance of sequences and sequence segmented vector （SSV） are put forward, and three theorems including sequence segmented mean theorem are proved. SSV-based whole sequence matching algorithm is designed to solve the whole match problem of number trend sequences. The algorithm uses radlans to measttre the trend, takes advantage of time of the trend maintenance to weight the value of trend, and realizes quick whole sequence similarity search of number trend sequences.