Brain Stroke Prognosis: a Fusion of Machine Learning and Deep Learning

Since brain stroke is one of the world's major causes of disability and death, accurate prediction models are essential for prompt diagnosis and treatment. The research analyzes combined approaches between machine learning (ML), deep learning (DL), and ensemble-based techniques for improving brain stroke outcome prediction. The research utilized a dataset containing demographics alongside lifestyle factors and medical information processed through normalization and imputation as well as the Synthetic Minority Oversampling Technique (SMOTE) which addressed class imbalance issues and solved missing values. The research examines four popular modeling approaches starting with Support Vector Machines (SVM) and Random Forest (RF) followed by CatBoost then Convolutional Neural Networks (CNN) and an improved ensemble combination. The ensemble model performed exceptionally well across comprehensive performance metrics which demonstrated accuracy at 96 % and precision at 97 % with recall at 96 %. Support Vector Machines showed outstanding recall performance which makes the model an ideal choice for medical applications that need successful diagnosis detection. The author stresses how ensemble approaches use different models to generate precise predictions for stroke outcomes. Advanced predictive systems in healthcare receive implementation guidance through these findings that will lead to enhanced patient results in critical medical conditions.