Predictive Machine Learning Techniques for Diabetes Detection: An Analytical Comparison

Diabetes is a syndrome that happens due to high blood glucose/sugar which is the major supplier of energy and is gained from the consumption of foods. When sufficient insulin is not produced by the pancreas, it results in diabetes. Till now there is no accurate solution to this disease and precaution can only avoid its side effects. It is mandatory to detect the disease at early stages so that side-effect on the patient remains quite low. Because of the advancement in technology now it is possible to identify the disease at its initial stages. This article uses various machine-learning techniques which can be applied to the dataset for making predictions that a person is suffering from diabetes or not. In this article, the data is collected from Kaggle and various machine-learning techniques such as Naïve Bayes, Logistic Regression, KStar, and Random Forest are employed on the dataset by using K fold cross-validation. Further, the performance of these techniques is evaluated on the basis of accuracy, f1-measure, precision, and recall, which determined that Random Forest performs better and shows the highest accuracy of 96.53%. Additionally, the proposed work may contribute to providing assistance to medical practitioners for the early diagnosis of diabetes.