Early diabetes risk classification using supervised learning algorithms

Abstract

Diabetes is one of the most devastating diseases and affects many people. Diabetes can be caused by a variety of causes, including ageing, obesity, inactivity, genetics, a poor diet, high blood pressure, and others. Diabetes increases the likelihood of developing several illnesses, including heart disease, renal disease, stroke, eye problems, nerve damage, etc. The information needed to diagnose diabetes is currently gathered through a variety of tests used in hospitals, and the diagnosis is then used to determine the best course of treatment. The healthcare sector has a considerable application for machine learning (ML). Databases in the healthcare sector are very vast. Big datasets can be examined using ML techniques to find hidden information and patterns, allowing one to learn from the data and predict outcomes properly. Using the existing methods, the forecast’s accuracy is not very good. In this study, we proposed an early diabetes prediction model that incorporates several extrinsic characteristics that contribute to the development of diabetes together with more widely used measures like polyuria, weight loss, polyphagia, visual blurring, alopecia, obesity, etc. The Support Vector Machine (SVM), the Logistic Regression (LOR), the Boosted Tree (BOT), and the Bagged Tree (BAT) are four different classifiers that are utilized in this paper to predict diabetes early on. The device’s performance is assessed in terms of accuracy, recall, specificity, precision, and f-measure. Results show that among the classifiers, BAT has the highest accuracy, at 98%.