Integrating SMOTE and LIME Techniques for Enhanced Stroke Prediction using Machine Learning Approaches

Stroke, a medical condition that occurs when there is a disruption in blood flow to the brain, leading to damage to brain cells. It is typically caused by either a blockage in a blood vessel (ischemic stroke) or a blood vessel rupture (hemorrhagic stroke). This study offers a holistic approach to addressing class imbalance in machine learning, specifically assessing the usefulness of the Synthetic Minority Over-sampling Technique (SMOTE) in conjunction with Z-scaling or standardization on a dataset. Various machine learning models, namely Logistic Regression, Support Vector Classifier, Naive Bayes, Random Forest, AdaBoost Classifier, XGB Classifier, K-nearest Neighbors, CatBoost Classifier, and Gradient Boosting Classifier, are evaluated using LIME explainable AI for model interpretability and analyzing various performance metrics. Notably, Random Forest demonstrates superior performance relative to other models, achieving an accuracy of 94.70%. This research contributes valuable insights for the development of diagnostic tools for enhancing stroke prediction.