Kernel Difference-Weighted k-Nearest Neighbors Classification

Abstract

Nearest Neighbor (NN) rule is one of the simplest and most important methods in pattern recognition. In this paper, we propose a kernel difference-weighted k-nearest neighbor method (KDF-WKNN) for pattern classification. The proposed method defines the weighted KNN rule as a constrained optimization problem, and then we propose an efficient solution to compute the weights of different nearest neighbors. Unlike distance-weighted KNN which assigns different weights to the nearest neighbors according to the distance to the unclassified sample, KDF-WKNN weights the nearest neighbors by using both the norm and correlation of the differences between the unclassified sample and its nearest neighbors. Our experimental results indicate that KDF-WKNN is better than the original KNN and distance-weighted KNN, and is comparable to some state-of-the-art methods in terms of classification accuracy.