Background Existing research focuses on the identification of key risk factors for stroke, improves the accuracy of stroke risk prediction, and provides more evidence for the scientific diagnosis, treatment and intervention of stroke.
Methods We included 4785 cases, stratified by sex (men: n=3156; women: n=1629) and age (18-40 years; 41-54 years; 55-69 years; 70+ years), the survey data of stroke patients conducted by Cooperative Hospital from 2019 to 2020. After data preprocessing, an extreme learning machine was used to construct a stroke risk prediction model.
Results The stroke risk prediction model was identified that total cholesterol and high-density lipoprotein were the 10 most important risk factors affecting the onset of stroke. The prediction accuracy rate of the risk prediction model is 97%.
Conclusion The method in this paper can quickly and effectively dig out the key risk factors affecting the onset of stroke from the data, and predict the risk of the onset, which has high application value.