Komparasi Jarak Euclidean dan Manhattan Pada Algoritma K-Nearest Neighbor Dalam Mendeteksi Penyakit Diabetes Mellitus

Abstract

Diabetes Mellitus is a chronic disease. This disease is caused by an increase in blood sugar levels in the body, it can cause diseases such as heart disease, obesity, and eye, kidney, and nerve diseases. Detection of Diabetes Mellitus is usually carried out by laboratory tests, so that patients have to undergo several medical tests to provide input values to a computerized diagnostic system which has proven to be expensive and has long queue times. From these problems, an artificial intelligence system is needed to diagnose this disease more easily and quickly. Therefore, the researcher aims to use an intelligent system to produce the highest accuracy from the results of the classification test using the K-Nearest Neighbor (K-NN) method with Euclidean distance and Manhattan distance. The class classifications used were pregnancy calculations, blood sugar in blood, blood pressure, skin fold thickness, insulin, body weight, diabetes genealogy dysfunction, and age. The research data in the form of datasets amounted to 450 datasets and the data was divided into two to determine the highest accuracy of 80% test data and 20% for training data. The highest accuracy using Euclidean distance is 84% with a value of K=5, and secondly, the Manhattan distance has the highest accuracy of 82% with a value of K=7.