基于点分布特征的多元时间序列模式匹配方法

多元时间序列模式匹配的常用方法难以刻画序列的全局形状特征,比如,Euclid方法的鲁棒性不够强;而PCA方法不适合处理小规模多元时间序列.基于点的统计分布提出了一种能够有效刻画多元时间序列形状特征的模式匹配方法.首先,提取多元时间序列样本的局部重要点,作为模式描述的方式;然后,根据重要点的统计分布特点构建特征模式向量,并借助Euclid范数来度量两个特征模式向量之间的相似程度,进而进行多元时间序列模式匹配.采用该方法进行模式匹配,充分利用了序列的全局形状特征.实验结果表明,基于点分布特征的多元时间序列模式匹配能够有效地刻画序列的形状特征,且能处理多种规模的序列数据.

Pattern Matching Method Based on Point Distribution for Multivariate Time Series

Common methods for matching multivariate time series such as the Euclid method and PCA method have difficulties in taking advantage of the global shape of time series. The Euclid method is not robust, while the PCA method is not suitable to deal with the small-scale multivariate time series. This paper proposes a pattern matching method based on point distribution for multivariate time series, which is able to characterize the shape of series. Local important points of a multivariate time series and their distribution are used to construct the pattern vector. To match pattern of multivariate time series, the Euclid norm is used to measure the similarity between the pattern vectors. The global shape characteristic is used in the method to match patterns of series. The results of experiments show that it is easy to characterize the shape of multivariate time series with this method, with which various scales can be dealt with in series data.