Predicting Early Risk of Diabetes using Hierarchical Fuzzy Logic

In recent years, diabetes has become a common disease that has affected people worldwide. Many factors make people have diabetes, like heavy body weight, increased cholesterol levels, lack of physical activity, bad food habits, family history, and so on. People with diabetes for a long time can get several complications like heart disorders, kidney failure, nerve injury, foot ulcer, diabetic retinopathy, etc. However, if the disease is predicted early, the risk of body complications can be reduced and people can live healthy for long periods. In this context, we have put effort into developing an early prediction model for diabetes using the Fuzzy Logic model and implemented it through MATLAB. The human reasoning rules are formulated based on Mamdani's principle taking three inputs; FPG, 2h-PG, and HbA1c. We have conducted extensive experiments with real-time data of diabetes patients collected from a local area in Hyderabad, India to analyze the model's performance. Further, we have developed a Hierarchical Fuzzy Inference System (FIS) called Fuzzy Tree to calculate the risk factor. The performance of the Fuzzy Logic and FT models are compared. The result of the study suggests that if the input features of the human body are selected correctly, the proposed models may predict the risk of diabetes proficiently and appropriately. However, the Fuzzy Tree provides a better output than the Fuzzy Logic model.