Machine Learning for Stroke Detection: Performance Evaluation of Random Forest and XGBoost Classifiers

Résumé fourni par la source

Stroke is the primary causes of death, posing a considerable amount of health and economic burden in lower income countries. Timely medical intervention and improved clinical results depend on early diagnosis and prediction. This research utilizes different machine learning models for stroke risk prediction based on different health related features which represent the symptoms observed in the patients. The dataset was cleaned by removing outliers and missing values and dropping the column containing null values. Feature importance analysis showed that the average glucose level and BMI were two of the strongest contributing features. Both models were hyperparameter-optimized using RandomizedSearchCV. Hypertuned Random Forest outperformed the rest with accuracy 99.85%.