A Hybrid Machine Learning Model in Diagnosing Brain Strokes

—Strokes can occur suddenly and unexpectedly, especially brain strokes, which can be fatal for individuals over the age of fifty. Survivors of a stroke may experience severe paralysis or weakness, posing a significant challenge for healthcare professionals to treat. However, artificial intelligence and Machine Learning (ML) have been proven promising in addressing these critical issues. Despite the high incidence of strokes in countries like Qatar, there is limited research on stroke risk in the Middle East. This study is the first to use a dataset that combines multiple open-source datasets from the region. In this research, several machine learning and ensemble learning algorithms, including Decision Trees (DT), Multiple Layer Perceptron (MLP), K-Nearest Neighbors (KNN), Gaussian Naïve Bayes (GNB), Logistic Regression (LR), Support Vector Machine (SVM), ensemble stacking, and Random Forest (RF) classifier have been investigated. All the algorithms were comprehensively analyzed by tuning their respective hyperparameters using the Grid Search approach and extracting the best features from the dataset through statistical analysis. The proposed ensemble stacking model achieved the highest accuracy and an F1-Score of 98% and 98.29%, respectively. The outcome indicates substantial improvement compared to current approaches in literature with similar datasets.