DocumentCode :
1383862
Title :
On the edited fuzzy K-nearest neighbor rule
Author :
Yang, Miin-Shen ; Chen, Chien-Hung
Author_Institution :
Dept. of Math., Chung Yuan Christian Univ., Chung Li, Taiwan
Volume :
28
Issue :
3
fYear :
1998
fDate :
6/1/1998 12:00:00 AM
Firstpage :
461
Lastpage :
466
Abstract :
Classification of objects is an important area in a variety of fields and applications. In the presence of full knowledge of the underlying joint distributions, Bayes analysis yields an optimal decision procedure and produces optimal error rates. Many different methods are available to make a decision in those cases where information of the underlying joint distributions is not presented. The k nearest neighbor rule (k-NNR) is a well-known nonparametric decision procedure. Many classification rules based on the k-NNR have already been proposed and applied in diverse substantive areas. The edited k-NNR proposed by D.L. Wilson (1972) would be an important one. Fuzzy theory, originated by L.A. Zadeh (1965), is widely used to represent the uncertainty of class membership. The fuzzy k-NNR has been proposed by several investigators. In this paper an edited type of the fuzzy k-NNR is developed. Next, some asymptotic properties of the proposed edited fuzzy k-NNR are created. Moreover, numerical comparisons are made between the proposed edited fuzzy k-NNR and the other fuzzy k-NNR. Those results confirm that the edited fuzzy k-NNR has a better performance than the fuzzy k-NNR
Keywords :
genetic algorithms; learning (artificial intelligence); pattern classification; simulated annealing; Bayes analysis; asymptotic properties; edited fuzzy K-nearest neighbor rule; k-NNR; nonparametric decision procedure; objects classification; optimal decision procedure; optimal error rates; Convergence; Councils; Extraterrestrial measurements; Fuzzy sets; Mathematics; Nearest neighbor searches; Pattern recognition; Uncertainty;
fLanguage :
English
Journal_Title :
Systems, Man, and Cybernetics, Part B: Cybernetics, IEEE Transactions on
Publisher :
ieee
ISSN :
1083-4419
Type :
jour
DOI :
10.1109/3477.678652
Filename :
678652
Link To Document :
بازگشت