Identification of glucose levels in urine based on classification using k-nearest neighbor algorithm method

Abstract

Abstract Glucose monitoring carried out through the urine testing to make it easier for patients to check their blood sugar without having to physically injure themselves and to prevent external bacteria from entering the body, which happens while using needles. This study aims to classify glucose-containing urine specimens based on diabetes levels by using the K-nearest neighbor method. Classification of urine specimens is achieved by using the Benedict method to produce the color of the urine specimen and the AS7262 sensor to detect the color produced by the specimen. The results showed that the classification of data on urine specimens has an accuracy of 96.33%. Previous studies conducted this experiment using a photodiode sensor and a TCS sensor, which produced red, green, and blue (RGB) colors. For identifying the color of a specimen, the AS7262 sensor can produce six colors (red, green, blue, yellow, violet, and orange) to identify the glucose level.