Title page for etd-0730113-152814


[Back to Results | New Search]

URN etd-0730113-152814
Author Chiech-an Tai
Author's Email Address No Public.
Statistics This thesis had been viewed 5571 times. Download 0 times.
Department Computer Science and Engineering
Year 2013
Semester 1
Degree Master
Type of Document
Language English
Title An Automatic Data Clustering Algorithm based on Differential Evolution
Date of Defense 2013-08-16
Page Count 63
Keyword
  • automatic clustering
  • data clustering
  • high-dimensional dataset
  • histogram analysis
  • differential evolution
  • Abstract As one of the traditional optimization problems, clustering still plays a vital role for the re-searches both theoretically and practically nowadays. Although many successful clustering algorithms have been presented, most (if not all) need to be given the number of clusters before the clustering procedure is invoked. A novel differential evolution based clustering algorithm is presented in this paper to solve the problem of automatically determining the number of clusters. The proposed algorithm, called enhanced differential evolution for automatic cluster-ing (EDEAC), leverages the strengths of two technologies: a novel histogram-based analysis technique for finding the approximate number of clusters and a heuristic search algorithm for
    fine-tuning the automatic clustering results. The experimental results show that the proposed algorithm can not only determine the approximate number of clusters automatically, but it can also provide an accurate number of clusters rapidly even for high dimensional datasets com-pared to other existing automatic clustering algorithms.
    Advisory Committee
  • Chu-Sing Yang - chair
  • Tzung-Pei Hong - co-chair
  • Chun-Wei Tsai - co-chair
  • Ming-Chao Chiang - advisor
  • Files
  • etd-0730113-152814.pdf
  • Indicate in-campus at 99 year and off-campus access at 99 year.
    Date of Submission 2013-09-13

    [Back to Results | New Search]


    Browse | Search All Available ETDs

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