A fuzzy K-NN algorithm using weights from the variance of membership values

In this paper, a new fuzzy K-nearest neighbor (K-NN) algorithm, called "Variance Weighted Fuzzy K-NN", is proposed. The main idea of this method is in giving weights to neighbors according to the standard deviation of their class membership values which reflect the value of a discriminant function. The classification results of 32 classes of complex images are given. Compared to the K-NN and fuzzy K-NN algorithms, our method shows an improved classification rate for various conditions.

[1]  Richard O. Duda,et al.  Pattern classification and scene analysis , 1974, A Wiley-Interscience publication.

[2]  A. Jówik,et al.  A learning scheme for a fuzzy k-NN rule , 1983 .

[3]  Keinosuke Fukunaga,et al.  The 2-NN Rule for More Accurate NN Risk Estimation , 1985, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[4]  Nicholas Kalouptsidis,et al.  Nearest neighbor pattern classification neural networks , 1994, Proceedings of 1994 IEEE International Conference on Neural Networks (ICNN'94).

[5]  Jack Koplowitz,et al.  The weighted nearest neighbor rule for class dependent sample sizes (Corresp.) , 1979, IEEE Trans. Inf. Theory.

[6]  James M. Keller,et al.  A fuzzy K-nearest neighbor algorithm , 1985, IEEE Transactions on Systems, Man, and Cybernetics.

[7]  Sahibsingh A. Dudani.  The Distance-Weighted k-Nearest-Neighbor Rule , 1976, IEEE Transactions on Systems, Man, and Cybernetics.

[8]  Keinosuke Fukunaga,et al.  Estimation of Classifier Performance , 1989, IEEE Trans. Pattern Anal. Mach. Intell..

[9]  Anil K. Jain,et al.  NOTE ON DISTANCE-WEIGHTED k-NEAREST NEIGHBOR RULES. , 1978 .