Predicting serious diabetic complications using hidden pattern detection

Diabetes is known as a metabolic disease. Type 1 diabetes is an anti-immune disease whereby the body's immune system kills off its own insulin producing beta cells in the pancreas. Type 2 diabetes is an advanced state of health in which the body becomes opposed to the usual impacts of insulin and/or progressively loses the capacity to produce adequate amount insulin in the pancreas, and it finally may not be able to produce any insulin. Type 2 diabetes is the most common form of the disease with complications including heart, vision, and foot conditions. This study aimed at predicting some important complications of Type 2 diabetes such as heart disease, retinopathy, diabetic foot, neuropathy, and nephropathy in order to provide useful information for patients. In this paper, in addition to introduce verity of sophisticated features for predicting Type 2 diabetics complications, a new dataset aiming at predicting Type 2 diabetics complications has been collected. To the best of our knowledge, the gathered dataset is only dataset for Type 2 diabetics' complications. Furthermore, several well-known classification algorithms have been trained and tested over a new dataset including 1,000 patients in order to introduce the best algorithms for detecting diabetic complications, including heart disease, retinopathy, diabetic foot, neuropathy, and nephropathy. The results indicate that the Random Forest algorithm is the best algorithm to predict the complications in most cases while the worst is the Naïve Bays classifier. The study findings can be useful to develop information systems in the field of health as a decision support system for physicians.