• DocumentCode
    1612236
  • Title

    Conflict evidence combination based on evidence classification strategy

  • Author

    Yanfei Chen ; Xuezhi Xia ; Yu An ; Shun Ge

  • Author_Institution
    Coll. of Comput. Sci. & Technol., Harbin Eng. Univ., Harbin, China
  • fYear
    2013
  • Firstpage
    167
  • Lastpage
    170
  • Abstract
    Evidence classification strategy is an effective method of conflict evidence combination. In order to improve the accuracy of the classification of evidence, this paper proposes an evidence classification strategy based on the degree of not subordination of evidences. Firstly, calculate all the degree of not subordination of evidences in the evidence set, eliminate the evidences with high degree of not subordination, and form a new evidence set, after that, eliminate the conflict evidences in the new set using recursive method, ensure that each classified evidence sets only have the low conflict evidences. Finally, combine evidences in every classification using Dempster´s rule, then compute the weights of all the classifications according to the evidence number in the classification, and get the final result by the weighted combination method. Simulation results show that the evidence combination based on this strategy is rational and effective.
  • Keywords
    case-based reasoning; inference mechanisms; pattern classification; uncertainty handling; Dempster rule; conflict evidence combination; conflict evidence combination method; evidence classification strategy; evidence not-subordination degree; evidence set; low conflict evidences; recursive method; weighted combination method; Educational institutions; Manganese; Reliability; Silicon; Sun; Tumors; Classification Strategy; Conflict; Evidence combination; degree of not subordination of evidences;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Chinese Automation Congress (CAC), 2013
  • Conference_Location
    Changsha
  • Print_ISBN
    978-1-4799-0332-0
  • Type

    conf

  • DOI
    10.1109/CAC.2013.6775722
  • Filename
    6775722