摘要
为解决心肺复苏(cardiopulmonary resuscitation,CPR)预测模型中性能无法满足临床应用要求,缺乏可解释性等问题,提出一种改进的堆叠融合CPR预测模型。首先采用人工少数类过采样算法(synthetic minority over-sampling technique,SMOTE)将数据均衡处理,嵌入法(Embed)进行特征筛选,其次使用4种树模型进行堆叠融合,五折交叉验证网格搜索算法进行参数优化,最后引入模型解释算法(shapley additive explanation,SHAP)量化特征重要性。结果表明,该模型具有良好的预测效果,准确率、精确率、召回率和F1分数分别达到了0.91、0.93、0.94和0.93,得到了CPR总时间、肾上腺素总剂量、年龄和既往疾病等重要影响因素。预测结果具有可解释性,可作为有效的辅助诊断模型。
In order to solve the problems of the cardiopulmonary resuscitation(CPR) prediction model in which the performance cannot meet the requirements of clinical applications and lacks interpretability,an improved stacked fusion CPR prediction model is proposed.Firstly,the synthetic minority over-sampling technique(SMOTE) is used for data balancing,and embedding is used for feature screening.Secondly,the four tree models are used for stacked fusion,and the five-fold cross-validation grid search algorithm is used for parameter optimization.And finally,shapley additive explanation(SHAP) is introduced as the interpretation algorithm to quantify feature importance.The results showed that the model had good prediction results with an accuracy of 0.91,precision of 0.93,recall of 0.94,and F1 score of 0.93,respectively,and obtained important influencing factor such as total CPR time,total epinephrine dose,age,and previous diseases.The prediction results were interpretable and can be used as an effective auxiliary diagnostic model.
作者
冯航测
田江涛
郝美林
孙洁
张瑛琪
Feng Hangce;Tian Jiangtao;Hao Meilin;Sun Jie;Zhang Yingqi(School of Electrical Engineering,North China University of Science and Technology,Tangshan 063210,China;Information Technology Center,China Mobile Communications Group Hebei Co.,Ltd.,Shijiazhuang 050000,China;Department of Emergency Medicine,First Hospital of Hebei Medical University,Hebei Province Emergency Technology Innovation Center,Shijiazhuang 050031,China)
出处
《国外电子测量技术》
北大核心
2023年第9期155-161,共7页
Foreign Electronic Measurement Technology
基金
2021年度河北省财政厅政府资助临床医学人才培养项目(LS202104)
2020年河北省科技计划(20477703D)
2020年度河北省财政厅老年病防治项目(LNB202010)资助。