Enhancing Diabetes Diagnosis: A Comparative Analysis of Machine Learning Algorithms and Evalution Metrics

Abstract

In recent years, artificial intelligence methods have been widely employed in various fields, including medical diagnosis applications. The main objective of this research is to implement a medical decision support system based on fuzzy decision trees and the PSO algorithm for diabetes diagnosis. To achieve this goal, a group-wise classification method is initially used to determine regression coefficients, which are then optimized using the particle swarm optimization (PSO) method. The PIMA dataset available on the UCI website is used to obtain the dataset. Additionally, MATLAB software is utilized for result simulation, and the results of the proposed method are compared with group-wise classification of numerical data, PSO optimization algorithm, and some algorithms used in the research background. The comparison metric is accuracy. The result obtained from applying the proposed method on the dataset used in this study yields an accuracy of 0.9033, indicating the superiority of the proposed method over GMDH and PSO algorithms, as well as other compared methods.