Study On Patients With Heart Failure Based On Logistic Regression Model

The incidence rate of cardiovascular and cerebrovascular diseases is the highest in chronic diseases,and the number of deaths due to heart failure is high.Analyzing the data of patients with heart failure and establishing the corresponding prediction model can provide reference for doctors to treat patients,provide decision support for hospital managers,and help hospitals allocate medical resources more effectively and reasonably.At present,there are two mainstream methods to predict the survival state of patients,one is to use traditional statistical methods;the other is to use machine learning method to simplify the variable parameters of prediction model,which can effectively solve the problem of insufficient prediction accuracy of traditional methods.This paper mainly solves the problem of predicting the survival status of patients with heart failure.This paper presents a prediction model based on the logistic regression algorithm.In the procedure of data pretreatment,in order to unify the measurement standard of data and eliminate the influence of units,the data is standardized.According to the binary classification scenario,we choose three algorithms: Logistic Regression(LR),Support Vector Machine(SVM)and Light Gradient Boosting Machine(LGB)to build the model.By comparing the fitting effect of the model,we select the optimal model as Logistic Regression.In order to solve the problem of data skew caused by unbalanced samples,we resample training data and balance data categories.In the variable selection stage,by comparing the effect of filtering and embedded variable selection methods to fit the model,we finally choose the embedded method to select variables,that is,according to the importance of variables in the model,the variables are gradually deleted,and the redundant information is removed.In this paper,the parameters of the model are optimized by grid search.The accuracy of the prediction model is verified by using the method of hold-out.70% of the original data set is used as training data and 30% is the test data.The results show that the F1 value is 0.8070,the accuracy is 87.78%,the precision is 88.46%,and the AUC value is 0.9136.This paper provides a modeling method with high accuracy,strong robustness and fast classification for predicting the survival state of patients with heart failure.