MACHINE LEARNING APPLICATION FOR STROKE PREDICTION

Abstract

Stroke is one of the leading causes of mortality and long-term disability worldwide, highlighting the need for effective early risk prediction tools. This study aims to develop a machine learning-based system for predicting stroke risk using structured clinical and demographic data. The Kaggle Brain Stroke Prediction dataset was utilized, and several classification algorithms were implemented, including Logistic Regression, Random Forest, XGBoost, LightGBM, and Voting Classifier. Data preprocessing techniques such as categorical encoding and Synthetic Minority Over-sampling Technique (SMOTE) were applied to address class imbalance. Model performance was evaluated using accuracy, precision, recall, F1-score and confusion matrix. Experimental results indicate that Logistic Regression combined with SMOTE achieved the highest recall, making it suitable for medical screening purposes, while ensemble models demonstrated higher overall accuracy but lower sensitivity. The findings emphasize the importance of handling imbalanced data and selecting appropriate evaluation metrics in healthcare prediction tasks.