Stroke Risk Prediction: Comparing Different Sampling Algorithms

Stroke is a serious disease that has significant impact on the quality of life and safety patients. Accurately predicting stroke risk great significance for preventing treating stroke. In past few years, machine learning methods have shown potential in risk. However, due to imbalance data challenges feature selection model selection, prediction still faces some difficulties.This article aims compare performance differences between different sampling algorithms prediction. This study used over-sampling algorithm (Random Over Sampling SMOTE), under-sampling Under ENN), hybrid (SMOTE-ENN), combined them with common such as K-Nearest Neighbors, Logistic Regression, Decision Tree Support Vector Machine build model.Through analysis experimental results, found SMOTE LR showed good prediction, high F1 score. addition, this overall undersampling better than oversampling algorithms.These research results provide useful references foundation further application. Future can continue explore more algorithms, methods, engineering techniques improve accuracy interpretability promote its application clinical practice.