Enhanced Stroke Risk Prediction with XGBoost Classifier: A Machine Learning Approach

Stroke remains a significant global health challenge, necessitating accurate risk prediction models for effective prevention and management. This study adopts a transfer learning approach to improve the prediction of stroke risk using the XGBoost classifier, which is strong and reliable. Since stroke datasets naturally tend to be imbalanced between class types, the SMOTE procedure was used during the preprocessing stage to reduce the bias in the stroke and non-stroke distributions. The study aims to achieve good prediction results based on demographic, lifestyle, and medical history data. Analytically, the research steps through data preprocessing through SMOTE, model building with XGBoost classifier, and thorough assessment using precision, recall, and F1-score metrics. The study highlights the importance of the XGBoost classifier in diagnosing the risk of stroke more efficiently than conventional approaches. For instance, the model offered an accuracy of 93.05 percent in identifying the preferred communication medium. Furthermore, the work defines potential predictors that are significant in increasing stroke incidence, which is useful in the context of the establishment of individual prevention measures. The present work makes significant steps towards the further development of the concept of individual-specific medical treatment of diseases and public health programs after a stroke.