Application of Adaptive Neuro-Fuzzy Inference System for diabetes classification and prediction

Abstract

Diabetes is one of the most common metabolic diseases and the statistics show that one in eleven adults has diabetes, but one in two adults with diabetes is undiagnosed, and in 2040 one in 10 adults will have diabetes. In this paper is proposed a hybrid Adaptive Neuro-Fuzzy Inference System (ANFIS) model for classifying patients with diabetes based on data sets with diabetic patients (Pima Indians Diabetes Dataset). The Pima Indians Diabetes Dataset contains 768 samples. In order to set the features vector of this system is used Diabetes Pedigree Function to define the fuzzy rule base with multiple premises. The Neuro-Fuzzy ANFIS modeling was implemented using ANFIS Fuzzy Logic Toolbox and MATLAB Toolbox. The performances of the algorithm were analyzed in terms of specificity, precision and sensitivity. The proposed neural network was trained and tested on Pima Indians Diabetes Database, proving an accuracy of 85.35% for training data and 84.27% for testing data.