Ensemble Machine Learning for Functional Outcome Prognostication in Acute Ischemic Stroke: A Multiclass Classification Study Using Clinical Risk Factors

DOI

No Doi

Published In

Year 2026 · Volume 1 · Issue 1

Published Online

January-April 2026

Pages

08-12

Article Type

Original Article

Check for updates

Background and Purpose: Acute ischemic stroke (AIS) imposes a substantial burden of mortality and functional disability worldwide. Accurate early prognostication of functional outcomes remains a critical clinical challenge. This study aimed to develop and evaluate multiclass supervised machine learning (ML) models for predicting 30-day (mRS 30) and 90-day (mRS 90) functional outcomes in AIS patients using routinely available clinical risk factors. Methods: A retrospective cohort of 146 AIS patients admitted to a private hospital, India, was enrolled. Fourteen clinical features including demographic, biochemical, and neurological parameters were extracted. Seven supervised ML algorithms — Logistic Regression, K-Nearest Neighbors, Decision Trees, Random Forest, XGBoost, Naive Bayes, and Kernel Support Vector Machine — were systematically evaluated. Class imbalance was addressed via synthetic minority oversampling (SMOTE/ADASYN). Model performance was assessed using mean testing accuracy, F1-score, and mean Area Under the Receiver Operating Characteristic Curve (AUC-ROC) across 10 iterations.Results: XGBoost demonstrated superior predictive performance, achieving a mean AUC of 0.66 for mRS 30 and 0.79 for mRS 90. Ensemble methods consistently outperformed single-algorithm classifiers. Non-linear classifiers surpassed linear models, indicating complex, non-linear interactions among AIS risk factors.