基于MW-REF算法的心肺复苏影响因素分析

针对传统模型对心肺复苏结果预测准确率较低、模型可解释性较差，提出了一种基于多模型加权递归消除法(MW-REF)的心肺复苏结果预测模型，并在Shapley加法解释(Shapley additive explanation, SHAP)框架下分析影响心肺复苏结果的关键因素。采用了随机森林、GBDT、XGBOOST作为基模型，将其特征重要性得分加权后使用递归消除法过滤特征并对3种及模型采用Voting进行模型融合，利用五折交叉验证下的准确率作为最终特征选择标准。最后对最终特征数据集下的融合模型进行可解释性分析。实验结果表明，与传统的递归特征消除算法对比，该模型提升了心肺复苏结果预测的准确率，模型预测结果具有可解释性，可为临床诊断提供辅助，提高诊断效率与心肺复苏成功率。

Analysis of influencing factors of cardiopulmonary resuscitation based on MW-REF algorithm

Aiming at the low prediction accuracy and poor interpretability of the traditional model for the outcome of cardiopulmonary resuscitation (CPR), a CPR outcome prediction model based on multi-model weighted recur-sive feature elimination (MW-RFE) was proposed, and the key factors affecting the outcome of CPR were an-alyzed under the framework of SHAP. Random forest, GBDT and XGBOOST were used as base models, and recursive elimination method was used to filter features after weighted feature importance scores. The three models and models were fused by Voting, and the accuracy under five-fold cross validation was used as the final feature selection criterion. Finally, the interpretability analysis of the fusion model under the final feature da-taset was conducted. The experimental results show that compared with the traditional REF algorithm, the model improves the accuracy of cardiopulmonary resuscitation prediction. The prediction results of the model are interpretable, which can provide assistance for clinical diagnosis and improve the efficiency of diagnosis and the success rate of cardiopulmonary resuscitation.