An Optimized Ensemble Bagging Model for mRS Score Prediction in Post-Stroke Patients

Stroke is a high-disability disease, with the modified Rankin Scale widely used to assess functional status and predict prognosis. This study aims to explore the relationship between 90-day mRS scores and factors such as patient history, medical history, onset information, and imaging results to predict functional recovery. First, Linear Discriminant Analysis (LDA) was applied for dimensionality reduction to extract key features, reducing redundancy and enhancing model efficiency. Multiple models were then constructed, including Decision Tree, Random Forest, Support Vector Machine, Logistic Regression, and Linear Regression. An Optimized Ensemble Bagging (OEB) model was proposed for prediction, with performance evaluated and compared using metrics such as F1-score and accuracy. Results showed that dimensionality reduction significantly improved model performance, with the OEB model achieving the best results. This study provides data support for stroke prognosis and identifies factors influencing mRS scores. Future research will focus on validating the predictive performance with larger sample sizes and diverse patient groups.