关键词: 多元时间序列, 分段, 相似性搜索, 特征点, 分层

Abstract: A variety of methods for matching multivariate time series can not measure similarity accurately and efficiently at the same time. This paper proposes a similarity search algorithm by segmentation based on special points of multivariate time series. It extracts the feature points, by which it segments the multivariate time series, and then transforms them into pattern sequences, thus the global shape characteristics of the original sequences can be retained. It makes the similarity search with the hierarchical matching method. The experimental results show that this method can effectively portray the global shape features of the sequences, retain local similarity matching by hierarchical matching, and improve the accuracy of the search at the same time.

Key words: multivariate time series, segment, similarity search, feature points, hierarchy