Comparison of LightGBM, SVM, and Logistic Regression Algorithms in Predicting Stroke Disease

Abstract

Stroke is a serious condition that can lead to disability or death due to disrupted blood flow to the brain. This study aims to compare three machine learning algorithms: LightGBM, Support Vector Machine (SVM), and Logistic Regression, in predicting the risk of stroke. The dataset used contains 5110 rows with 12 attributes, including demographic information and health history. The research process began with data preprocessing, followed by splitting the data into training and testing sets. Models were then trained using the three algorithms and evaluated using accuracy, precision, recall, and F1-score metrics. The analysis results indicate that Logistic Regression performed the best overall, providing a balance between detecting stroke cases and identifying healthy individuals. SVM showed stable results with a balance between recall and precision, while LightGBM, despite high accuracy, was less effective in detecting stroke cases. The study concludes that Logistic Regression is the most suitable model for predicting stroke risk, though SVM can be a good alternative.