基于卷积神经网络和Adaboost的心脏病预测模型

Heart disease prediction model based on convolutional neural network and Adaboost

借助计算机技术,使用年龄、性别等基本特征预测心脏病的易感性,对心脏病的早期预测和防治具有重要意义。针对基于机器学习的心脏病预测模型准确率不高的问题,提出一种基于卷积神经网络(CNN)和Adaboost的心脏病预测模型CNN-Adaboost。首先,对原始数据进行预处理,结合特征相关性与特征组合算法融合两两属性特征,并升维数据,使各属性特征充分融合;然后,通过CNN进行充分的特征提取;最后,结合Adaboost机器学习算法建立心脏病预测模型。UCI数据集上的测试结果表明,CNN-Adaboost预测模型优于K近邻(KNN)等传统机器学习模型和K近邻-随机森林(KNN-RF)等优化模型,准确率、AUC、查准率和查全率可达到0.917、0.95、0.924与0.85。CNN-Adaboost模型具有良好的分类效果,能为医患人员进行心脏病预测与预防提供帮助。

With the help of computer technology,predicting susceptibility to heart disease using basic features such as age and gender is important for early prediction and prevention of heart disease.In order to address the problems of low accuracy of heart disease prediction models based on machine learning,a heart disease prediction model CNN-Adaboost was proposed based on Convolutional Neural Network(CNN)and Adaboost.Firstly,the original data was pre-processed,and the two-two attribute features were fused by combining feature correlation and feature combination algorithms,then the data was up-dimensioned to fully fuse each attribute feature.Secondly,the full feature extraction was performed by using CNN.Finally,the heart disease prediction model CNN-Adaboost was constructed by using Adaboost machine learning algorithm.Experimental results on the UCI dataset show that the proposed CNN-Adaboost prediction model outperforms traditional machine learning models such as KNN(K Nearest Neighbor)and optimization models such as KNN-RF(Random Forest)with accuracy of 0.917,AUC of 0.95,precision of 0.924 and recall of 0.85,respectively.The CNN-Adaboost model has a good classification effect,which is a feasible and effective method to help doctors and patients in heart disease prediction and prevention.