Subsequence searching is a non-trivial task in time series data analysis and mining. In recent years, different approaches are published to improve the performance of subsequence searching which based on index the time series and lower bound the Euclidean distance. In this paper, the problem of applying Euclidean distance on time series similarity measure is first reviewed. Previous approaches to align time series for similarity measure are then adopted for subsequence searching, they include: dynamic time warping (DTW) and perceptually important point (PIP). Furthermore, a tree data structure (SB-Tree) is developed to store the PIP of a time series and an approximate approach is proposed for subsequence searching in the SB-Tree. The experimental results performed on both synthetic and real datasets showed that the PIP approach outperformed DTW. The approximate approach based on SB-Tree can further improve the performance of the PIP-based subsequence searching while the accuracy can still be maintained.
|Name||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Conference||3rd International Conference on Fuzzy Systems and Knowledge Discovery, FSKD 2006|
|Period||24/09/06 → 28/09/06|
- Theoretical Computer Science
- Computer Science(all)