Weighted K-nearest Centroid Neighbor Classification

Abstract

The k-Nearest Centroid Neighbor rule (KNCN), as an extension of the k-Nearest Neighbor rule (KNN), is one of the promising algorithms in pattern classification. In this article, we take into consideration the proximity and spatial distribution of the neighbors by means of nearest centroid neighborhood for a query pattern, and introduce two weighted voting schemes for KNCN. Experimental results show that the proposed classifiers are effective algorithms, and obtain much improvement over the state-of-the-art KNN based algorithms. 1553-9105/