Diabetics Prediction in Real Life Data Using Machine Learning Algorithm

Abstract

There are many diseases in Bangladesh where diabetes is ordinary disease of human body. A human is affected when his body sugar level over a period is pretty much high. It also causes of stroke, heart attack, kidney failure and blindness etc. It is possible to control if you understand the earlier stage and can save a human life. This is a disease which is in growing day by day. The motive of this study to do the measurement of the performances of some popular Machine Learning algorithms. In recent years, Machine Learning is a wonderful platform which has a huge impact on different corner of science and technology including medical sectors. There are many algorithms in Machine Learning. But in this paper work, we have used five popular ML algorithms which is Gaussian Naïve Bayes, Random Forest, Support Vector Machine, Logistic Regression and Decision Tree to find out the measurement of performances. These algorithms have been trained and tested on real data for diabetes patients in Bangladesh. The performances of these techniques are enlisted. We have used diabetes patient dataset, conducted in 2021, derived from Islami Bank Hospital and Diagnostic Center. The dataset consists of 485 data which contains 267 true class and the rest of data is false class. When we applied algorithms then we find that Random Forest based algorithm gives 97% accuracy. So, the RF based classifier is better than other algorithms.