Diagnosing Diabetes using Artificial Intelligence

Abstract

Diabetes is a condition that causes high blood sugar level in body. If sugar levels aren’t maintained properly, then diabetes is a disease where it causes damage to the other parts of body. If diagnosing and treatment of diabetic patients are postponed, it will lead to some major issues such as heart attacks, renal failure etc. Most important problem within the medical world is timely and precise diagnosis of disease. Issues of patience can be reduced if the disease is diagnosed at the proper time and special medical aid is provided to the patients. In this study, PIDD data set is used and different models are implemented. Different model includes Decision tree, Support vector machine, Logistic regression, K neighbor classifier, XGboost, Ensemble Hybrid [Naïve baye, KNN, Logistic regression], Multilayer perceptron, Naïve baye, Random forest. The accuracies of those models range between 67% to 78%. Proposed model is implemented using recurrent neural network [RNN] and it provides accuracy of 90.4%. The results shows that proposed RNN model provides higher accuracy level i.e. 90.4% than the prevailing models and other models.