Kernel-based fuzzy K-nearest-neighbor algorithm

In this paper, the fuzzy k-nearest-neighbor is extended to a kernel-based model which performs a nonlinear classification by kernel methods. This generalized model is called kernel-based fuzzy k-nearest-neighbor model. Through a nonlinear mapping the input data are mapped into a high-dimensional feature space where fuzzy k-nearest-neighbor is performed. The computation of the nonlinear mapping is finished implicitly by kernel methods. The kernel methods are used as a chief means of computing fuzzy k-nearest-neighbor efficiently in high-dimensional feature space where the nonlinear pattern now appears linear. The effectiveness of the proposed algorithm is shown for classification in application to the real world data sets. The proposed model compares favorably with fuzzy k-nearest-neighbor.