• DocumentCode
    3110215
  • Title

    A Novel Classifier Selection Approach for Adaptive Boosting Algorithms

  • Author

    Ali, ABM Shawkat ; Dobele, Tony

  • Author_Institution
    Sch. of Comput. Sci., Central Queensland Univ., Rockhampton, QLD
  • fYear
    2007
  • fDate
    11-13 July 2007
  • Firstpage
    532
  • Lastpage
    536
  • Abstract
    Boosting is a general approach for improving classifier performances. In this research we investigated these issues with the latest Boosting algorithm AdaBoostMl. A trial and error classifier feeding with the AdaBoostMl algorithm is a regular practice for classification tasks in the research community. We provide a novel statistical information- based rule method for unique classifier selection with the AdaBoostMl algorithm. The solution also verified a wide range of benchmark classification problems.
  • Keywords
    learning (artificial intelligence); pattern classification; AdaBoostMl algorithm; adaptive boosting algorithm; benchmark classification problem; machine learning; novel classifier selection approach; statistical information-based rule method; Australia; Boosting; Decision trees; Electronic mail; Error analysis; Machine learning; Machine learning algorithms; Performance evaluation; Testing; Training data;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Computer and Information Science, 2007. ICIS 2007. 6th IEEE/ACIS International Conference on
  • Conference_Location
    Melbourne, Qld.
  • Print_ISBN
    0-7695-2841-4
  • Type

    conf

  • DOI
    10.1109/ICIS.2007.38
  • Filename
    4276436