The Impact of the Distance Weighting Function on the Performance of KNearest Neighbor Algorithm in Medical Data sets

Abstract

The K nearest neighbor (KNN) technique is well recognized and extensively used in the field of machine learning classification algorithms. In this study, the performance of distance weighting functions, one of the most important factors affecting the performance of KNN, was compared. These functions are equal, inverse, and squared-inverse. The performance of the functions was examined in five different medical data sets. To evaluate the performances, a confusion matrix and five different metrics commonly used to evaluate classification problems were used. Out of the five data sets used in the research, the inverse KNN method yielded effective outcomes in four of them, while both the equal and squared-diverse methods achieved success in three data sets.