摘要
基于机器学习技术及信息增益特征的评估方法,文章采用MPI技术所得到的患者心脏参数及医生诊断的临床数据构建冠心病诊断支持模型,为冠心病诊断提供了一种智能决策方法。通过对江苏省XX医院脱敏真实临床数据的筛选,得到同时进行过“双源CT”以及“核素”检查的病人共621例,对其补全及离散化处理后随机分成521例的训练集及100例的测试集。采用信息增益方法进行特征选择以筛选出对冠心病诊断支持的相关特征,并使用不同机器学习技术构建分类模型。在逻辑回归、随机森林、决策树构建的冠心病诊断模型中,通过信息增益最终筛选出Gender,LHR,SRS等14个特征后构建的决策树分类诊断模型在测试集准确率达到72%,该决策树算法训练出的模型最适合用于该冠心病数据集“双源CT”检查结果的预测。
Based on machine learning technology and information gain feature evaluation method,using MPI technology to obtain the patient’s heart parameters and clinical data diagnosed by the doctor to build a coronary heart disease diagnosis support model to provide an intelligent decision-making method in this paper.A total of 621 patients who underwent both“dual-source CT”and“nuclides”examinations were screened from the clinical data of desensitization from XX Hospital of Jiangsu Province.After completion and discretization,they were randomly divided into 521 training sets and 100 test sets.The information gain method was used to select features to screen out relevant features supporting the diagnosis of coronary heart disease,and different machine learning techniques were used to construct classification models.In the coronary heart disease diagnosis model constructed by logistic regression,random forest,and decision tree,the decision tree classification and diagnosis model constructed after 14 features such as Gender,LHR,and SRS are finally filtered by information gain.The accuracy rate of the test set reached 72%.The model trained by the decision tree algorithm is most suitable for the prediction of the“dual-source CT”test result of the coronary heart disease data set.
作者
姚妮
高政源
王强
朱付保
Yao Ni;Gao Zhengyuan;Wang Qiang;Zhu Fubao(School of Computer and Communication Engineering,Zhengzhou University of Light Industry,Zhengzhou 450000,China)
出处
《无线互联科技》
2020年第9期135-138,共4页
Wireless Internet Technology
基金
河南省2020年科技发展计划项目,项目编号:202102210384
郑州轻工业大学2019年众创空间孵化项目,项目编号:2019ZCKJ228。
关键词
机器学习
冠心病诊断
特征选择
决策树
逻辑回归
machine learning
coronary heart disease diagnosis
feature selection
decision tree
logistic regression
