A MACHINE LEARNING FRAMEWORK FOR EARLY PREDICTION OF BRAIN STROKE USING CLINICAL ATTRIBUTES

Stroke remains one of the leading causes of mortality and long-term disability worldwide, necessitating early detection for timely clinical intervention and improved patient outcomes. Inspired by methodologies used in ADMETbased drug side-effect prediction—where functional group patterns and engineered descriptors facilitate early risk assessment—this study proposes an explainable machine learning framework for predicting stroke occurrence based on clinical, demographic, and lifestyle factors. The curated dataset incorporates key attributes such as age, hypertension status, history of heart disease, average glucose level, smoking behavior, and body mass index (BMI). Comprehensive data preprocessing techniques, including exploratory data analysis (EDA), feature correlation analysis, and class imbalance handling using the Synthetic Minority Over-sampling Technique (SMOTE), were employed to enhance data quality and model robustness. Multiple machine learning models, including Logistic Regression, Random Forest, Gradient Boosting, and Extreme Gradient Boosting (XGBoost), were developed and evaluated. Among these, the Gradient Boosting model demonstrated superior and consistent generalization performance across both balanced and imbalanced datasets. Furthermore, SHapley Additive exPlanations (SHAP) analysis was utilized to interpret model predictions, revealing that age, average glucose level, hypertension, and BMI are the most influential features—aligning with established clinical risk factors for stroke. The proposed framework provides a reliable, interpretable, and data-driven approach for early stroke risk prediction. By leveraging structured feature engineering concepts analogous to functional group-based modeling in drug discovery, this work highlights the potential of machine learning in enhancing predictive accuracy in healthcare diagnostics. The system can assist clinicians in identifying high-risk individuals and lays the foundation for future integration of multimodal biomedical data to further improve stroke prediction models.