Diagnosis of Diabetes Mellitus using Adaptive Neuro-Fuzzy Inference System

Abstract

Extensive research efforts have been put into improvising the clinical decision support systems to diagnose diabetes in terms of accuracy and interpretability. However, optimization was still a matter of concern. Diabetes being a long-lasting health disorder requires precise diagnosis to administer optimal dosage of insulin and medication on a regular basis. Fuzzy based systems have the tendency to handle vagueness and uncertainty in data set of any size which is suitable for decision making in diabetes diagnosis vis-a-vis medical domain. Fuzzy system limitation in adjusting during a learning process was overcome by implementing ANFIS which is an integration of the advantage of fuzzy control interpolation and adaptability through neural network back propagation. The current research presents ANFIS implementation for better classification accuracy with minimum error. Initially, fuzzy based system was designed for the Pima Indian diabetic database followed by a well-trained and tested ANFIS implementation to classify results efficiently

Keywords-Clinical Decision Support Systems (CDSS); Adaptive Neuro-Fuzzy Inference System (ANFIS); Root Mean Square Error (RMSE); Fuzzy Inference System (FIS)