Towards Transparent and Accurate Diabetes Prediction Using Machine Learning and Explainable Artificial Intelligence
Pir Bakhsh Khokhar, Viviana Pentangelo, Fabio Palomba, Carmine Gravino·January 30, 2025
Summary
A study introduces a framework for diabetes prediction using machine learning (ML) models with explainable artificial intelligence (XAI) tools. The ensemble model identifies BMI, age, general health, income, and physical activity as key predictors, achieving high accuracy. The study addresses the need for both accuracy and interpretability in ML predictions, focusing on local interpretable model-agnostic explanations (LIME) and SHapley Additive explanations (SHAP). It aims to fill a gap in systematically evaluating XAI algorithms for diabetes-related predictions, building upon state-of-the-art XAI tools within a healthcare context. The study emphasizes the importance of explainability and robustness in AI systems for diabetes prediction, offering a framework for incorporating explainability tools into medical settings to support lifestyle modifications and reduce diabetes risk.
Introduction
Background
Overview of diabetes prevalence and management challenges
Importance of accurate and interpretable predictive models in healthcare
Objective
To introduce a framework for diabetes prediction using ML models with XAI tools
To identify key predictors for diabetes using ensemble models
To evaluate and compare XAI algorithms for diabetes-related predictions
Method
Data Collection
Description of the dataset used for diabetes prediction
Data sources and collection methods
Data Preprocessing
Data cleaning and normalization
Feature selection process
Model Development
Ensemble model selection and training
Key predictors identified through the model
XAI Tools Evaluation
Implementation of LIME and SHAP
Evaluation criteria for XAI algorithms
Model Evaluation
Metrics for assessing model accuracy and interpretability
Results
Key Predictors
Detailed analysis of BMI, age, general health, income, and physical activity
Model Performance
Accuracy and interpretability of the ensemble model
XAI Algorithm Comparison
Comparative analysis of LIME and SHAP
Discussion
Importance of Explainability
Role of explainable AI in healthcare decision-making
Robustness and Accuracy
Balancing model accuracy and interpretability
Framework for Incorporation
Practical considerations for integrating XAI into medical settings
Conclusion
Summary of Findings
Recap of the framework's contributions to diabetes prediction
Future Directions
Potential areas for further research and development
Implications for Practice
Recommendations for healthcare professionals and policymakers
Basic info
papers
software engineering
machine learning
artificial intelligence
Advanced features
Insights
What is the main focus of the study mentioned in the text?
Which factors does the ensemble model identify as key predictors for diabetes prediction?
How does the study aim to address the gap in systematically evaluating XAI algorithms for diabetes-related predictions?
What are the two XAI tools highlighted in the study for interpreting the ML model's predictions?