Diabetes Diagnostic Prediction Using Vector Support Machines

The most important factors for the diagnosis of diabetes mellitus (DM) are age, body mass index (BMI) and blood glucose concentration. Diagnosis of DM by a doctor is complicated, because several factors are involved in the disease, and the diagnosis is subject to human error. A blood test does not provide enough information to make a correct diagnosis of the disease. A vector support machine (SVM) was implemented to predict the diagnosis of DM based on the factors mentioned in patients. The classes of the output variable are three: without diabetes, with a predisposition to diabetes and with diabetes. An SVM was obtained with an accuracy of 99.2% with Colombian patients and an accuracy of 65.6% with a data set of patients of a different ethnic background.