Explainable Machine Learning for Diabetes Risk Prediction
AI Research

Explainable Machine Learning for Diabetes Risk Prediction

Transparent AI models that provide clear, human-understandable insights for early diabetes risk detection.

Overview

This research develops machine learning models for Type-2 diabetes prediction that are both accurate and explainable, enabling clinicians and individuals to understand and trust AI-driven decisions.

?The Challenge

Diabetes management is a daily challenge that affects lifestyle, diet, and long-term wellbeing. Many AI models achieve strong accuracy but operate as black boxes, limiting their use in healthcare where trust and clarity are essential.

Our Solution

We designed interpretable models supported by explainability techniques such as feature importance and LIME. These provide clear reasoning behind predictions, helping clinicians and patients make informed decisions.

Key Features

Feature importance and rule-based explanations
Clinician-friendly visual insights
Patient-understandable explanations
Uncertainty-aware predictions
Comparison with clinical risk scores
Interactive model exploration

Results & Impact

  • 1Published in IEEE IMCOM 2026
  • 2High predictive performance with strong accuracy
  • 3Improved trust in AI-driven predictions
  • 4Supports safer clinical decision-making

Technical Approach

Models tested include Random Forest, XGBoost, LightGBM, and CatBoost. Explainability integrated using LIME and feature importance, ensuring alignment with clinical reasoning.

Future Directions

Focus on adaptive explanations and real-world clinical validation for deployment.

Technologies Used

Explainable AIClinical MLLIMEHealthcare AI

Interested in similar work?

We're always open to discussing new partnerships and research collaborations.

Get in touch