Brain Stroke Predictions Considering Optimized Ensemble Machine Learning Techniques

Abstract

Brain stroke is a potentially fatal medical emergency that occur when the blood flow to the brain is disrupted or diminished, depriving brain tissue of oxygen and nutrients. It is a primary contributor to mortality and chronic disability globally. Timely identification of stroke risk is essential for prompt intervention, which can markedly enhance results and preserve lives. Conventional statistical models have been utilized for stroke prediction, depending on established assumptions on the relationships among variables. This study seeks to examine the correlation among general health, blood pressure, and the incidence of cerebral stroke using a thorough methodology. Ensemble learning, a method that integrates numerous base models to enhance prediction accuracy, has demonstrated significant potential in healthcare applications. Existing ensemble approaches possess drawbacks, particularly in their inability to adequately manage skewed datasets. The amalgamation of metaheuristic optimization methods with ensemble learning facilitated the development of resilient and precise models. Several machine learning (ML) techniques, including Random Forest (RF), Support Vector Machine (SVM), Logistic Regression (LR), Gradient Boosting (GB), Stacking Classifier (SC), Boosting Classifier (BC), and Voting Classifiers (VC), are being studied for their ability to predict stroke risk. Following a rigorous examination that takes into account several performance factors such as recall, accuracy, F1 score, and precision, the Stacking Ensemble technique is determined to be a superior choice to current stroke detection methods, with an accuracy of 91.57%.