Mining Closed Patterns in Time-Series Databases
Date Issued
2010
Date
2010
Author(s)
Wu, Huei-Wen
Abstract
Closed pattern mining is a critical research issue in the area of knowledge discovery and data mining with the aim of discovering interesting patterns hidden in a large amount of data. In this dissertation, we propose three algorithms, called CMP (Closed Multi-sequence Patterns mining), CFP (Closed Flexible Patterns mining), and CNP (multi-resolution Closed Numerical Patterns mining) to solve various issues extended from the problem of mining closed patterns.
The CMP algorithm is designed to find closed patterns in a multi-sequence time-series database. The CFP algorithm is developed to solve the problem of mining closed flexible patterns in a time-series database. Both the CMP and CFP algorithms involve a transformation of time-series sequences into symbolic sequences in the first phase. Although analyzing on symbolic sequences is ideal to reduce the effect of noises and ease the mining process, these approaches may lead to pattern lost and the sequences supporting the same pattern may look quite different.
To overcome the problem raised in symbolic sequence analysis, the CNP algorithm is proposed to mine closed patterns without any transformation from time-series sequences to symbolic sequences. The method also employs the Haar wavelet transform to discover patterns in the multiple resolutions in order to provide different perspectives on datasets.
All the proposed algorithms have employed the concept of projected databases to localize the pattern extension that leads to a significant runtime improvement. Moreover, effective closure checking schemes and pruning strategies are devised respectively in each of the proposed algorithms to avoid generating redundant candidates.
The experimental results show that the CMP algorithm significantly outperforms the modified Apriori and BIDE algorithms. The CFP algorithm achieves better performance than the modified Apriori algorithm in all cases. And, the CNP algorithm has demonstrated a significant runtime improvement in comparison to the modified A-Close algorithm.
Subjects
Data mining
Closed pattern
Time-series database
File(s)![Thumbnail Image]()
Loading...
Name
ntu-99-D95725007-1.pdf
Size
23.32 KB
Format
Adobe PDF
Checksum
(MD5):6953249c6b3b083e9cb9bbaf4f579626