Asynchronous Periodic Pattern Mining for Cyclic and Incremental Sequential Time Stamp
Abstract
Mining of periodic patterns in time-series databases is an interesting data mining problem. It can be envisioned as a tool for forecasting and prediction of the future behavior of time-series data. Most researches focused on mining synchronous periodic patterns, but in practice some periodic patterns can not be recognized because of presence of random noisy and disturbance in large datasets. To increase the efficiency of mining Asynchronous periodic patterns on large datasets, our proposal work move in the direction of finding all maximal complex patterns in a single step algorithm using a single dataset scan without mining single event and multiple events patterns explicitly. The asynchronous periodic patterns are mined using depth first search technique. Three parameters are employed to specify the minimum number of repetitions required for a valid segment of non disrupted pattern occurrences, the maximum allowed disturbance between two successive valid segments, and the total repetitions required for a valid sequence.
To find multiple periods based partial periodic patterns, looping over single period using the hit set based approach is one of the naive methods. The algorithm is to directly apply the max-pattern hit set method to each period and the sequence. The proposal evaluates another method for multiple periods using shared mining. This method is similar to the max-pattern single period mining algorithm. During the first scan of the sequence for all periods, the frequent pattern and candidate max-pattern are generated. During the second scan the hit sets of all the periods are generated as in the second scan of max-pattern hit set method. The problem of max-sub pattern tree construction and derivation of frequent patterns from the max-sub pattern tree is also discussed.
The proposal of our work integrates the cyclic partial periodic patterns and incremental patterns with adaptive thresholds to produce interactive mining of partial periodic patterns. The maximum pattern search space and the execution overhead for the proposal work are analyzed with synthetic and real data sets from UCI Repository. In addition it work in the direction of merge mining a generalization of incremental mining which discover patterns of two or more databases that are mined independently of each other. An improved version of merge mining algorithm is planned to build for asynchronous partial periodic patterns in time-series databases.Keywords
Full Text:
PDFReferences
Ahmed, C.F. Tanbeer, S.K. Byeong-Soo Jeong Young-Koo Lee, Efficient Tree Structures for High Utility Pattern Mining in Incremental Databases, Knowledge and Data Engineering, IEEE Transactions on Dec. 2009 Volume: 21, Issue: 12
DFD Ding-An Chiang Cheng-Tzu Wang Shao-Ping Chen Chun-Chi Chen The Cyclic Model Analysis on Sequential Knowledge and Data Engineering, IEEE Transaction on Nov 2009
Ya-Han Hu, Tony Cheng-Kui Huang, Hui-Ru Yang and Yen-Liang Chen, On mining multi-time-interval sequential patterns, Science Direct Journal May 2009
J. Han, G. Dong, and Y. Yin Efficient mining partial periodic patterns in time series database 15th International Conference on Data Engineering,, pages 106–115, 1999.
Heikki Mannila, Hannu Toivonen, and A. Inkeri Verkamo Discovering frequent episodes in sequences
B. Ozden, S. Ramaswamy, and A. Silberschatz. Cyclic association rules. 14th International Conference on Data Engineering, pages 412–421, 1998.
R. Srikant and R. Agrawal. Mining sequential patterns: Generalizations and performance improvements. 5th International Conference on Extending Database Technology, volume 1057, pages 3–17. Springer, 1996.
K. H. Tung, J. Han H. Lu, and L. Feng. Breaking the barrier of transactions: Mining inter-transaction association rules, International Conference on Knowledge Discovery and Data Mining pages 297–301, 1999.
J. Yang, W. Wang, and P.S. Yu Mining asynchronous periodic patterns in time series data Sixth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pages 275– 279, 2000.
Refbacks
- There are currently no refbacks.
This work is licensed under a Creative Commons Attribution 3.0 License.