<![CDATA[Stroke occurs due to disrupted blood flow to the brain, either from a blood clot (ischemic) or a ruptured blood vessel (hemorrhagic), leading to brain tissue damage and neurological dysfunction. It remains a leading cause of death and disability worldwide, making early prediction crucial for timely intervention. This study evaluates the impact of data balancing techniques on stroke prediction performance across different machine learning models. Random Forest (RF) consistently achieves the highest accuracy (98%) but struggles with precision and recall variations depending on the balancing method. Decision Tree (DT) and K-Nearest Neighbors (KNN) benefit most from SMOTE and SMOTETomek, improving their F1-scores (11.21% and 9.18%), indicating better balance between precision and recall. Random Under Sampling enhances recall across all models but reduces precision, leading to lower overall predictive reliability. SMOTE and SMOTETomek emerge as the most effective balancing techniques, particularly for DT and KNN, while RF remains the most accurate but requires further optimization to improve precision and recall balance.]]>
Copyrights © 2024
