MBR-safe piecewise aggregate approximation for time-series subsequence matching

초록

In this paper we present an MBR-safe version of piecewise aggregate approximation (PAA) and propose an efficient subsequence matching method based on that MBR-safe PAA. PAA is known as an excellent lower-dimensional transformation due to its simple computation and performance improvement in similar sequence matching. The MBR-safe transformation is known to reduce the number of lower-dimensional transformations in similar sequence matching since it can transform a high-dimensional minimum bounding rectangle (MBR) itself to a low-dimensional MBR directly. Thus, by integrating these advantages of PAA and MBR-safety, we propose an MBR-safe version of PAA and improve subsequence matching performance using that transformation. Contributions of the paper can be summarized as follows. First, we propose a PAA-based MBR-safe transformation, called mbrPAA, and formally prove its MBR-safety and optimality Second, we propose an mbrPAA-based subsequence matching method and formally prove its correctness. Third, we present the concept of entry reuse property, and using this property we propose an efficient way of constructing high-dimensional MBRs in subsequence matching. Fourth, we show the superiority of mbrPAA through extensive experiments. Experimental results show that, compared with the previous approach, our mbrPAA is 24.2 times faster in constructing low-dimensional MBRs and improves subsequence matching performance by up to 65.9%.

키워드