Analyzing the Probability of Diabetes Through Machine Learning

Abstract

Numerous diseases impact people around the world on a daily basis, worrying many about their physical health and well-being. Throughout the world, diabetes is one of the most widespread diseases, affecting hundreds of millions of individuals. Machine learning may be able to diagnose patients with diabetes based on their medical records. Using different models such as Logistic Regression, Random Forests, and Neural Networks, we have found that it is possible to predict the probability of having diabetes. The neural network had the highest AUROC (Area Under the Receiver Operating Characteristic curve), AUPRC (Area Under the Precision-Recall Curve), and Accuracy of 0.8330, 0.4316, and 0.8597, respectively, making it the best-performing model out of the three. The results of this paper suggest that machine learning models, specifically neural networks, may be useful in diabetes diagnosis.