ML_Diabetes: An Efficient Framework for Diabetes Predictions Using Machine Learning Techniques

Diabetes is a chronic disease that can be fatal; currently, millions suffer from it. The diagnosis process can be tedious, long, or even inaccurate; in addition, the delay in diagnosis can cause serious health problems that include amputation, kidney failure, blindness, or even death. Thus, measures should be taken to ease the process of diagnosis. In this paper, the aim is to propose a methodology to apply machine learning and neural network models to predict diabetes based on a set of features that are applied to different data sets. In addition, those models are proposed to speed up the diabetes diagnosis process and predict diabetes accurately, as this will ease the process for doctors and patients. Moreover, we shed light on other research paper's shortcomings concerning diabetic prediction. This paper utilized eight models, seven of which are machine learning models, including Support Vector Machine, Decision Tree, Naive Bayes, Gradient Boosting, K nearest Neighbors, Random Forest, and Logistic Regression, while one model is a Neural Network. The following results were established: Neural Network achieved very high accuracies in the three tested data sets: 82.3%, 91.9%, and 75.5% for Pima Indian, Diabetes 2, and Mendeley diabetes data-sets, respectively. Gradient Boosting, Logistic regression, Support vector Machine, Naive Bayes weren't far behind, but they sometimes underperformed. As shown in this paper, machine learning has immensely helped in the prediction of diabetes, but with even more future improvements could be closer to 100 %.