Pendeteksian Kadar Glukosa dalam Darah pada Gejala Diabetes Tipe 1 Menggunakan Algoritma K-Nearest Neighbor dengan Metode Nafas

Abstract

The goal of this study was to detect glucose levels in the blood using a non-invasive method through human mouth breath. In patients with type 1 mellitus generally have low salivary levels which can cause bad breath or called Halitosis. The method used in this study is using breath sensor in the form of MQ-4 and Figaro TGS-2602 on human mouth breath to get results in the form of hydrogen sulfide (H2S) and methane (CH4) from a person s breath. The results will be obtained in mg / dl after the data is obtained by a sensor with a filter the last lowpass was processed using a machine-learning algorithm in the form of KNearest Neighbor with the Regression classification method. The results of the 5 diabetes mellitus sample test data and 40 diabetes mellitus training data can detect glucose in the blood with an accuracy of 80% and will be compared with previous research. Sample 40 training data was taken from several patients who had p diabetes mellitus and non-diabetes mellitus disease using glucometer with 95% accuracy rate. This system is expected to provide a solution for people with type 1 diabetes mellitus for someone who suffers from the disease..