ESPM: A Model to Enhance Stroke Prediction with Analysis of Different Machine Learning Approaches and Hyperparameter Tuning

Stroke prediction is paramount in healthcare to enable timely intervention and reduce the burden of this devastating condition. This research paper examines the prediction of strokes using machine learning methods, aiming to enhance accuracy and efficiency in risk assessment. Numerous Machine Learning (ML) techniques, such as Support Vector Machine (SVM), XGBoost, Random Forest, Linear Regression, and Gaussian Naive Bayes, are explored using a comprehensive dataset containing patient demographics, medical history, lifestyle factors, and clinical measurements. Based on different ML models, an Enhanced Stroke Prediction Model (ESPM) is proposed. Grid search, Randomized search, and Bayesian optimization are employed as hyperparameter tuning techniques, and parameters like accuracy, precision, recall, and F1 score are analyzed. It is observed that SVM with Grid Search hyperparameter tunning performs well with an accuracy of 94.129%; Positive Predictive Value (PPV), True Positive Rate(TPR), and F1 Score achieved are 89%, 94%, and 91%, respectively. The outcomes demonstrate the suitability of these models for different aspects of stroke prediction, such as handling complex patterns, capturing non-linearity, robustness to noisy data, and modeling continuous risk scores.