Diabetes prediction using supervised machine learning

Abstract

Diabetes is a disease that can lead to blindness, kidney failure, and heart attacks, as well as death. According to the International Diabetes Federation, there were 463 million diabetics in 2019. If predictions are correct, this number will rise by 578 million by 2030, reaching 700 million by 2045. According to an article published by the Ministry of Health of the Republic of Indonesia in 2020, the ten countries with the highest diabetes rates in 2019 include Indonesia. The ability of experts is required to determine the type of diabetes disease. Because of their delay in discovering what disease they have, many people who are examined have a disease that can be described as severe. Diabetes detection technology is required to prevent severe conditions. In today's medical world, doctors can use it to quickly and accurately interpret diseases. Because of that we can use machine learning to prevent the death by making an artificial inteligent model that can predict diabetes disease and the method that be used is comparison between the KNN and Naive Bayes algorithms to see which algorithm suit the best for diabetes prediction. The study concluded by comparing two k-Nearest Neighbor algorithms and the Naive Bayes algorithm to predict diabetes based on several health attributes in the dataset using supervised machine learning. According to the results of our experiments and evaluating alghorithm using Confusion Matrix, the Naive Bayes algorithm outperforms KNN.