Stroke prediction using synthetic minority over-sampling technique and extreme gradient boosting

Area of Science:

• Biomedical Informatics
• Machine Learning in Healthcare
• Cardiovascular Disease Research

Background:

• Stroke is a leading global cause of death, necessitating accurate early prediction models.
• Interpretable models are vital for guiding stroke prevention and clinical care strategies.

Purpose of the Study:

• To develop and evaluate high-accuracy, interpretable machine learning models for early stroke risk prediction.
• To identify key predictors of stroke using advanced explainability techniques.

Main Methods:

• Utilized the Kaggle stroke dataset, applying SMOTE for class imbalance.
• Trained and compared XGBoost, Random Forest, LightGBM, CatBoost, and Support Vector Machine (SVM) models.
• Employed 10-fold cross-validation for robust performance evaluation and SHAP for model interpretability.

Main Results:

• XGBoost demonstrated superior performance with 97.26% accuracy, surpassing other models and prior benchmarks.
• The Shapley Additive Explanations (SHAP) algorithm identified age and hypertension/blood pressure as dominant stroke predictors.
• The study provided both global feature importance rankings and case-level predictive insights.

Conclusions:

• The developed pipeline offers a practical and interpretable approach to stroke risk prediction.
• Achieved state-of-the-art performance, making the model suitable for clinical decision support systems.
• Highlights the potential of machine learning in enhancing cardiovascular disease prevention and management.