Comparison of Random Search and Grid Search for Optimizing SVM in Stroke Disease Classification

This study examines the application of machine learning for stroke disease classification, emphasizing the need for precise model optimization to enhance predictive accuracy. The research focuses on the performance of the Support Vector Machine (SVM) algorithm, comparing two hyperparameter tuning approaches: Random Search and Grid Search. The methodology includes stages of Exploratory Data Analysis (EDA), Data Preprocessing, Classification Modeling, Performance Evaluation, and Hyperparameter Tuning using a stroke classification dataset. Results reveal that SVM optimized with Random Search achieved an accuracy of 90.44%, outperforming the Grid Search method. These findings highlight the significance of effective data preprocessing and model optimization in developing reliable predictions and advancing strategies for better stroke prevention and management. The study demonstrates the pivotal role of machine learning in early stroke detection and risk assessment, providing insights into practical applications for healthcare. The results establish that random search is preferred for improving SVM performance in critical diagnostic tasks.