Interpretable Machine Learning Approaches for Advancing Disease Prognosis

The integration of Multilayer Perceptron (MLP) neural networks and Explainable AI (XAI) techniques represents a significant advancement in disease prediction models. This paper proposes a novel framework that leverages MLPs combined with XAI methods, such as Local Interpretable Model-agnostic Explanations (LIME) and Partial Dependence Plots (PDP), to enhance both predictive accuracy and interpretability. Our approach improves upon existing studies that often utilize simpler models with limited datasets and lack transparency. By evaluating three significant health conditions—heart disease, diabetes and Parkinson's disease—this research achieves accuracies of 87.18% for Parkinson's, 86.89% for heart disease, and 75.97% for diabetes, while providing actionable insights into model predictions. This work not only demonstrates the feasibility of deploying AI-driven diagnostics in clinical settings but also establishes a foundation for future research in interpretable AI for healthcare applications.