Research And Application Of Auxiliary Diagnosis Algorithm For Chronic Kidney Disease Based On Machine Learning

Recently,machine learning technology has been more and more widely used in the medical field.Using the technology can help doctors find hidden diseases early and correctly predict disease progression,thereby reducing the morbidity and mortality of patients.Chronic kidney disease and its main type Immunoglobulin A(IgA)nephropathy are chronic disease with high incidence and poor prognosis.Exploring the key risk factors that affect chronic kidney disease and timely medical intervention for patients can help slow down or even reverse the course of the disease progressing.This thesis focused on the research on the auxiliary diagnosis algorithm of chronic kidney disease,mainly to solve the two problems of how to use machine learning technology to establish a more efficient auxiliary diagnosis model of chronic kidney disease and how to use machine learning technology to establish a prediction model of children’s IgA nephropathy.The main research results of this thesis are as follows:1.Propose a diagnosis model of chronic kidney disease based on the optimized XGBoost algorithm.After preprocessing the original dataset,the thesis used the random forest model to select features of the dataset,and finally selected five features that optimal the model classification performance;for parameter optimization problem of the XGBoost algorithm,propose a hybrid algorithm of genetic and particles swarm optimization;the optimized XGBoost algorithm was compared with the algorithms of support vector machine,extreme learning machine and K nearest neighbor.The results showed that the chronic kidney disease classification model based on optimized XGBoost had the best performance,and the accuracy,precision,recall,F1-score and AUC value were all 100%,which were improved compared with the latest research results.2.It is proposed to use the random forest algorithm to establish a predictive model of children’s IgA nephropathy.Through data analysis of the IgA nephropathy original dataset,it was found that the ratio of positive and negative samples was close to 1:14,so the thesis proposed to use the Borderline-SMOTE2 algorithm to process the unbalanced datasets.The chi-square test method was used to perform the feature correlation analysis on obtained balanced dataset,select the most relevant features,and then compared random forest with XGBoost,support vector machine,k-nearest neighbors and extreme learning machine algorithms for building prediction models of children’s IgA nephropathy.It could be concluded that the prediction model of children’s IgA nephropathy based on random forest algorithm had the best performance from the experimental result,with accuracy of 97.01%,precision of 99.22%,recall of 94.76%,F1-score of 96.93%,and AUC value of 98.58%.Use the Django framework in Python and HTML text language and base on the random forest model proposed to establish a prediction system of children’s IgA nephropathy,the patient enters the designated laboratory indicators to obtain the prediction results of whether the patient will reach end-stage renal disease within 5 years.