Identifying Risk Factors of Diabetes using Fuzzy Inference System

Abstract

Identification of the real risk factors of diabetes is still very much inconclusive. In this paper, fuzzy rules based system was devised to identify risk factors of diabetes. The system consists of five input variables: Body Mass Index, age, blood pressure, Creatinine, and serum cholesterol and one output variable: level of risk. Three Gaussian membership functions for linguistic terms are defined for each input variable. The level of risk is defined using three triangular membership functions to represent output variable. Based on the information from patients’ clinical audit reports, the system was used to classify the level of risk of fifty patients that currently undergoing regular diagnosis for diabetes treatment. The system successfully classified the risk into three levels of Low, Medium and High where three main contributing factors toward developing diabetes were identified.
The three risk factors are age, blood pressure and serum cholesterol.
The multi-input system that characterised by IF-THEN fuzzy rules provide easily interpretable result for identifying predictors of diabetes. Research to establish reproducibility and validity of the findings are left for future works.