Research On Heart Disease Prediction Method Based On Data Mining Classification Technology

In recent years,the death rate of heart disease has been increasing,and the age of heart disease patients has been decreasing,with some young and middle-aged people also at higher risk.Heart disease occurs when the heart has a problem and is unable to supply blood to the rest of the body.Therefore,how to effectively predict heart disease and achieve early detection and treatment has become a hot issue in the current medical field.Fortunately,as AI matures,researchers are applying data mining techniques to disease research,providing new ways to predict heart disease.Based on the classification technology in data mining,this thesis applies the corresponding algorithms to the prediction of heart disease research and experiment,and the results are compared and analyzed.The main research content of this thesis is as follows:(1)Research on coronary heart disease risk prediction based on enhanced logistic regression algorithmWith the gradual application of machine learning in the field of heart disease prediction,the accuracy rate of prediction is still need to be improved.Therefore,in order to improve the accuracy of prediction of heart disease,a coronary heart disease risk prediction model based on enhanced logistic regression algorithm was proposed in this thesis.Based on Ada Boost algorithm,logistic regression algorithm was used as a weak learner for multiple iterative training.Thus,the prediction model MEB-LR(Medical Ensemble learning-Logistic Regression)was constructed,and common single models were compared and verified with the models before and after enhancement.The experimental results show that the proposed model has higher prediction accuracy.(2)Heart disease prediction of Light GBM model optimized by IHB based on incomplete dataBecause the data of heart disease is often incomplete,and most hyperparameter optimization algorithms have the problem of poor optimization efficiency.Therefore,IHB-Light GBM(Improved HyperbandLight Gradient Boosting Machine)heart disease prediction model based on incomplete data is proposed in this thesis.Firstly,weight values are introduced on the basis of Hyperparameter sampling in Hyperband alg orithm to improve the parameter optimization ability of the algorithm.Secondly,K-nearest neighbor algorithm is used to interpolate missing values of incomplete data.Finally,the improved IHB optimization algorithm is used to optimize the global parameters of Light GBM,and the prediction model of heart disease is established.The results show that the parameter optimization effect of IHB algorithm is better than that of other common optimization algorithms,and the model in this thesis is significantly better than that of random forest and other evaluation indexes,which can obtain faster prediction speed and higher prediction accuracy.(3)Research on multi-classification algorithm of arrhythmia based on CNN-LSTMThe process of image feature extraction by machine learning is complex,the extraction ability is limited,and the electrocardiogram signal has noise interference.In this thesis,ECG signals are denoised by wavelet transform,and GAN network is used to enhance the ECG data after processing.Then,a Convolutional neural network(CNN)and Long Short Term Memory(LSTM)multi-classification model of arrhythmia is proposed,and the MIT-BIH ECG data set is used for experimental simulation.The experimental results show that compared with other methods,the proposed model has obvious advantages in the classification of arrhythmia,and provides useful value for related medical experiments.