基于Relief-F算法的心血管介入患者术后死亡风险预测

Prediction of postoperative death risk in patients with cardiovascular intervention based on the Relief-F algorithm

针对心血管介入患者全周期病理数据普遍存在缺失、不连续、非结构化等问题,建立了心血管介入专病数据库,并采用基于Relief-F算法的预测方法,对心血管介入患者术后死亡风险进行预测。首先参照HL7、CDISC等国际心血管疾病统一标准对各数据源进行标准化处理,建立研究数据集,并对数据进行清洗和预处理;其次采用Relief-F算法对特征进行选择,最终保留30个特征变量;再次选择逻辑回归、支持向量机、随机森林等3种机器学习方法进行建模分析,并采用10折交叉验证方法对分类器进行训练;最后引入准确率等模型评价指标来评估各算法在数据集上的分类预测效果。实验结果表明:随机森林的分类效果在该研究数据集上的表现最佳,准确率达到81.97%,精确率为86.90%,召回率为82.14%,F1值为0.8441。该研究提出的方法能够客观反映患者术后死亡风险,为心血管介入患者术后死亡风险预测提供了一种有效的解决方案。

In view of the common problems such as missing,discontinuous and unstructured pathological data of patients with cardiovascular intervention throughout the whole cycle,a cardiovascular interventional disease database was established,and the prediction method based on the Relief-F algorithm was adopted to effectively predict the risk of postoperative death of patients with cardiovascular intervention.Firstly,all data sources were standardized according to HL7,CDISC and other international cardiovascular disease standards to obtain research data sets,and the data sets were cleaned and preprocessed.Secondly,the Relief-F algorithm was used to select the features,and 30 feature variables were retained in the end.Thirdly,logistic regression,support vector machine and random forest were selected for modeling and analysis,and the 10-fold cross-validation method was used to train the classifier.Finally,model evaluation indexes such as accuracy rate were introduced to evaluate the classification prediction effect of each algorithm on the data set.The experimental results show that the classification effect of random forest has the best performance on the research data set,its accuracy rate is 81.97%,the accuracy rate is 86.90%,the recall rate is 82.14%,and the F_1 value is 0.8441.This study can objectively reflect the postoperative death risk of patients,and provides an effective solution for predicting the postoperative death risk of patients with cardiovascular intervention.