Multiclass classification scheme for diagnosis of diabetes mellitus based on type-1 fuzzy systems

Diabetes is a chronic disease when there is insufficient insulin produced by the pancreas or ineffective use of the insulin by the body, causing high blood sugar. There are three types of diabetes known as type 1, type 2 and, Gestational. The cause of this metabolic disease is not only the heredity or family history, but also due to the lack of knowledge, unawareness and carelessness that lead to the unhealthy diet practices and lifestyles. In this study, Mamdani fuzzy inference system is used to diagnose this disease among women in helping them to control and prevent from getting diabetes mellitus disease. This study aims to examine the symptoms of positive diabetes or negative diabetes and to ascertain the accuracy of the fuzzy system in diagnosing diabetes mellitus. All female patients represented in the dataset diagnosed with diabetes mellitus aged at least 21 years old from Pima Indian heritage living near Phoenix, Arizona. Graphical User Interface (GUI) application was developed using Fuzzy Logic Toolbox software and MATLAB to implement the models. The results from diagnosis of diabetes mellitus by using Mamdani Fuzzy Inference System method obtained an accuracy rate of 90% where 260 women with positive diagnoses and 76 with negative diagnoses. This demonstrates that Mamdani Fuzzy Inference System has the capacity and adaptability in managing this issue as well as a good alternative to diagnose diabetes.