Narges Shafaei Bajestani, Maryam Aradmehr, Ensieh Nasli Esfahani, Behrooz Khiabani Tanha,
Volume 18, Issue 2 (2-2019)
Abstract
Background: Diabetes is one of the most dangerous and common diseases of the modern world. Since medical research usually has limited data available and medical data is very ambiguous, it seems appropriate to use the fuzzy model to find out the relationship between input and output in medical data. None of the previous articles of fuzzy regression have been used to predict complications of diabetes, including nephropathy. Therefore, in this study, a fuzzy regression model was used to predict nephropathy in a diabetic patient.
Methods: In the present study, GFR results of previous patient experiments were used to predict a deeper horizons of GFR and ultimately to predict renal disease. Chronic kidney disease has been stratified based on the amount of GFR, that fuzzy data has been constructed based on these levels. The GFR prediction was performed in the following steps. Step 1: Define fuzzy sets based on the GFR level, which is considered for each level of a fuzzy set. Step 2: Fuzzify patient data Based on fuzzy sets. Step 3: GFR prediction with fuzzy regression model. Step 4: Defuzzifying the predictions. Step 5: Evaluating the model efficiency. The RMSE error is used to compare the performance of the model.
Results: The results of GFR prediction showed that comparison RMSE was 10.09 with using simple linear regression model and 4.24 in fuzzy model.
Conclusions: fuzzy regression model can predict nephropathy in diabetic patients.
