摘要
文章采用简单随机抽样(旁置法)的方式,按 7:3 的比例划分训练集与测试集.基于该训练集建立支持向量机、BP神经网络、决策树、随机森林、Adaboost算法、加权K-近邻等分类模型,利用测试集对心力衰竭死亡风险预测模型的效果进行测试.使用精度、查全率、查准率、卡帕系数(Kappa)、F1 分数等评价指标评判各种模型调优后的预测效果,最后选出BP神经网络为最佳的疾病风险预测模型,为临床医学研究诊断心力衰竭提供一些参考性意见.
Abstract
This paper uses Simple Random Sampling method,and divides the training set and the test set in a 7:3 ratio.Based on this training set,SVM,BP Neural Network,Decision Tree,Random Forest,Adaboost Algorithm,and Weighted K-nearest Neighbors classification models are established,and the test set is used to test the effect of the Heart Failure death risk prediction model.Accuracy,Recall Ratio,Precision Rate,Kappa coefficient,F1 score and other evaluation indexes are used to evaluate the prediction effect of various models after optimization.Finally,the BP Neural Network is selected as the best disease risk prediction model,which provides some reference opinions for clinical diagnosis of Heart Failure medical research.
国家科技期刊平台
NSTL
万方数据