An Empirical Evaluation of Machine Learning Models for Stroke Risk Assessment with Explainable AI

Stroke has significant socioeconomic effects, particularly in low- and middle-income nations, and is a major cause of mortality and disability worldwide. This study applies machine learning models to predict stroke, leveraging clinical and demographic data to highlight those who may be at risk. We applied a range of models, including Random Forest, XGBoost, Gradient Boosting, and Neural Networks, after comprehensive data preprocessing. Among these, Random Forest achieved the highest accuracy (98.87%) and ROC AUC (1.0), making it highly effective for stroke risk prediction. We improved transparency in healthcare applications by emphasizing the relevance of individual risk factors in forecasts using LIME for model interpretability.. These findings support the use of ensemble models and explainable AI for reliable, interpretable stroke prediction, paving the way for preventive healthcare.