Artificial intelligence application for diabetes prediction

Abstract

This study investigates multiple approaches for the prediction of type 2 diabetes Based on Biometric signs. Three supervised machine learning models (Logistic Regression, Random Forest, and XGBoost) were developed and evaluated based on their predictive accuracy, feature interpretability, and computational performance. Additionally a fuzzy logic system and a rule-based expert system approaches were implemented to simulate human reasoning and clinical decision-making. The models were applied to the Pima Indians Diabetes dataset and tested using a combination of statistical metrics and visual diagnostics. Results show that while machine learning algorithms outperform in terms of raw accuracy, fuzzy and expert systems offer greater transparency and explainability. This work highlights the complementary strengths of data-driven and rule-based systems in the design of intelligent diagnostic tools for healthcare.