<![CDATA[Background: Recurrent stroke is a major cause of increased morbidity and disability. Traditional scores, the ESRS indicates that this score has limited discriminatory ability. The study aims to develop a Machine Learning (ML) model based on individual ESRS components that can surpass the traditional score’s accuracy for stroke recurrence risk. Methods: The study employs a Comparative Cross-Sectional design, with a total of 115 data classified into First-Time Stroke and Recurrent Stroke. Preprocessing with SMOTE for class imbalance. The classification model was built using the Random Forest algorithm and validated with 10-Fold Stratified Cross-Validation in WEKA software. Results: The Optimal ML Model achieved a superior Area Under the Curve (AUC) of 0.949 (significantly exceeding the traditional ESRS AUC of 0.55–0.58), demonstrating extremely strong discriminatory capability. Conversely, the traditional ESRS Model showed very low Accuracy and poor Precision. Conclusion: The development of the Random Forest Machine Learning model based on individual vascular risk factor components proved to be significantly superior in assessing stroke recurrence risk compared to traditional ESRS risk stratification. With an AUC of 0.949. These findings justify the potential integration of the ML model into Clinical Decision Support Systems (CDSS).]]>
Copyrights © 2026
