基于机器学习与Shap可视化对冠心病患者死亡生存分析与可解释性分析

打开文本图片集

中图分类号：0343.1；O341 文献标志码：Adoi： 10.3969/j .issn.1673-5862.2025.04.010

A machine learning shap-based study on survival analysis interpretability for coronary heart disease patients

LIURuiyin，LIWei （College ）

Abstract： Cardiovascular disease remains a major global public health challenge，with heart failure in particular associated with high mortality rates.To enhance the prediction survival outcomes in heart failure patients，this study applies various machine learning classification algorithms in conjunction with SHAP-based interpretability analysis to systematically evaluate key features influencing the risk coronary heart disease. Based on clinical data from 299 heart failure patients collected in 2Ol5，the study identifies serum creatinine ejection fraction as the two most critical predictors survival，using biostatistical tests feature importance rankings during model training validation. Results show that models constructed using only these two features outperform those based on the full set variables in terms predictive accuracy，while also fering superior interpretability. These findings provide both theoretical technical support for earlydiagnosis personalized intervention in cardiovascular disease.

Key Words： cardiovascular diseases； shapley additive explantions； feature ranking； featureselection；machine learning

心血管疾病，包括冠心病、脑血管病和心力衰竭等，导致全球每年约1700万人死亡[1。（剩余5152字）