Machine Learning Approaches for Stroke Risk Prediction: Findings from the Suita Study
Problem
Stroke is a major public health concern, and predicting who will have one and identifying key risk factors from cohort data remains challenging.
Why it matters
Stroke drives substantial mortality and morbidity, so early risk identification enables prevention.
What the AI did
K-prototype clustering grouped 7389 Suita Study participants into risk clusters, and five supervised models (LR, RF, SVM, XGBoost, LightGBM) predicted stroke, with SHAP identifying key predictors such as age and systolic blood pressure.
Results
not stated in source
Impact
Random Forest performed best and SHAP identified age and systolic blood pressure among key predictors; no numeric metrics stated.
Record
| Venue | Journal of Cardiovascular Development and Disease |
|---|---|
| Published | 2024 |
| Work type | article |
| Authors | Thien Vu, Yoshihiro Kokubo, Mai Inoue, Masaki Yamamoto, Attayeb S. Mohsen, Agustin Martin Morales, Takao Inoué, Michihiro Araki |
| Citations | 30 (OpenAlex, at import time) |
| References | 45 |
| Model or method | Random Forest, XGBoost, LightGBM, SVM, Logistic Regression |
| Organization | not stated in source |
| Primary source | https://doi.org/10.3390/jcdd11070207 |
| Open access | gold |
| Keywords | stroke risk, random forest, SHAP, clustering, cohort study |
Importance rating
significant — thousands to millions affected
Abstract calls stroke a significant public health concern due to mortality and morbidity impact; risk prediction and biomarker identification enable prevention, though no numeric disease scale is given.