Paper Info
Title
Time series subsequence searching in specialized binary tree
Abstract
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.
Year
DOI
Venue
2006
10.1007/11881599_67
FSKD
Keywords
Field
DocType
pip-based subsequence,time series data analysis,euclidean distance,time series,dynamic time warping,specialized binary tree,approximate approach,time series subsequence,time series similarity measure,different approach,pip approach,previous approach,data structure,time series data,indexation,binary tree,lower bound
Longest increasing subsequence,Longest common subsequence problem,Dynamic time warping,Similarity measure,Computer science,Artificial intelligence,Pattern recognition,Euclidean distance,Tree (data structure),Algorithm,Binary tree,Subsequence,Machine learning
Conference
Volume
ISSN
ISBN
4223
0302-9743
3-540-45916-2
Citations 
PageRank 
References 
2
0.38
11
Authors
4
Name
Order
Citations
PageRank
Tak-chung Fu140721.29
Hak-pun Chan220.38
Fu Lai Chung3153486.72
Chak-man Ng41169.33