Classification of Hypertension Patients in Palembang by K-Nearest Neighbor and Local Mean K-Nearest Neighbor

Abstract

Classification is a multivariate technique for separating different data sets from an object and allocating new objects into predefined groups. Several methods that can be used to classify include the k-Nearest Neighbor (KNN) and Local Mean k-Nearest Neighbor (LMKNN) methods. The KNN method classifies objects based on the majority voting principle, while LMKNN classifies objects based on the local average vector of the k nearest neighbors in each class. In this study, a comparison was made on the results of classifying hypertensive patient data at the Merdeka Health Center in Palembang City with the KNN and LMKNN methods by looking at the accuracy and the smallest APER value produced. The results showed that by using the same proportion of training and testing data and choosing different k values, the results of classifying hypertension patient data at the Merdeka Health Center in Palembang City with the KNN and LMKNN methods resulted in the APER value or the same error rate and accuracy, namely sequentially equal to 0.0303 and 96.97%.