A weighting approach for KNN classifier

Abstract

In this paper, a weighting approach for k nearest neighbors (kNN) algorithm is proposed. The motivation of the proposed approach is to find the optimal weights via Artificial Bee Colony (ABC) algorithm. To test the validity of the hybrid algorithm called ABC based distance-weighted kNN, dW-ABC kNN, four UCI data sets (Iris, Haberman, Breast Cancer, and Zoo) are used. The results reveal that dW-ABC kNN algorithm improves the correct classification performance in Iris, Haberman, and Breast Cancer data set. The performance degradation occurs when it is applied on Zoo data set. It can be concluded that ABC algorithm is applicable to kNN algorithm.