Title page for etd-0802114-021426


[Back to Results | New Search]

URN etd-0802114-021426
Author Yu-Chieh Tseng
Author's Email Address No Public.
Statistics This thesis had been viewed 5569 times. Download 406 times.
Department Computer Science and Engineering
Year 2013
Semester 2
Degree Master
Type of Document
Language English
Title Online Fusion of Across-Store High Utility Patterns
Date of Defense 2014-07-28
Page Count 92
Keyword
  • utility mining
  • online data mining
  • on-shelf time period
  • multi-site environment
  • Data mining
  • Abstract Utility mining, which takes the quantities and profits of items in a set of transactions into consideration, has become an emerging research topic due to its wide applications. Most of the existing studies on utility mining, however, only consider data coming from a single database. In reality, large organizations usually have a chain of stores in different locations. Besides, how to flexibly response to different users’ query conditions from multiple data sources is also a big challenge. Accordingly, in this thesis, we develop appropriate online mining approaches for finding high utility itemsets in a multi-site environment. We first introduce a new research issue called online multi-site utility mining, which considers not only the quantities and profits of items in transactions, but also the time periods and locations of the items in a multi-site environment. We then propose a Three-Phase Online Multi-site Utility mining algorithm (abbreviated as TP-OMU) to efficiently and effectively find such patterns. A useful strategy to predict the upper-bounds of utility values of items in a multi-site environment is also designed to reduce the number of candidates in the mining process. We next extend the idea above to on-shelf consideration. Items in the chain stores may be put on the shelf and taken the off shelf multiple times. We thus introduce another research issue called online multi-site on-shelf utility mining, which considers not only the utilities of items, but also the on-shelf locations and time periods of the items in a multi-site environment. Meanwhile, an efficient mining method is developed as well to cope with the problem. Finally, the experiments on synthetic datasets are conducted to show the effectiveness of the two kinds of patterns and the performance of the two proposed approaches under different parameter settings. The results demonstrate that they can perform well in a multi-site environment.
    Advisory Committee
  • Chang-Shing Lee - chair
  • Ming-Chao Chiang - co-chair
  • Chien-Feng Huang - co-chair
  • Tzung-Pei Hong - advisor
  • Files
  • etd-0802114-021426.pdf
  • Indicate in-campus at 2 year and off-campus access at 2 year.
    Date of Submission 2014-09-02

    [Back to Results | New Search]


    Browse | Search All Available ETDs

    If you have more questions or technical problems, please contact eThesys