Research On Nephropathy Assistant Diagnosis System Based On Machine Learning

At present,the auxiliary diagnosis of nephropathy solved by machine learning in China is mostly based on TCM syndrome type.However,there are few cases of assistant diagnosis of nephropathy in western medicine based on blood biochemistry and urine routine tests.Moreover,the diagnosis of nephropathy of TCM syndrome type is mostly whether or not the kidney disease is present.There is little research on the clear diagnosis of kidney diseases such as kidney failure and kidney stones in machine learning.This paper combines machine learning with medical information technology.This article takes the common Western medicine kidney disease test data as experimental data.The three classification algorithms of KNN,decision tree and random forest were used to establish a classification model of nephropathy to diagnose six common kidney diseases.In view of the lack of diagnostic performance of the three models,this paper proposes a solution to solve the problem of multi-classification of six common kidney diseases by using two-class classification.And using the random forest algorithm to conduct experiments.The diagnostic accuracy of the new model for nephropathy was 87.7%.The research content and research focus of this paper are as follows:(1)Explain the flow of the experiment in this paper by introducing the workflow of machine learning in medical applications.Using the functional classification of machine learning in the medical field leads to the idea of using the classification algorithm to solve the diagnosis of common kidney disease.According to the theory of common classification techniques of machine learning,the modeling scheme of this experiment is proposed.(2)The source and data characteristics of the experimental data are introduced.In view of the irregularity,redundancy and missing of data sets,the data are preprocessed.Six kinds of multi-classification models of nephropathy were established by using k-nearest neighbor,decision tree and random forest.Theexperimental results show that the random forest classification model of nephropathy has the highest accuracy rate of 63.7%.(3)In order to improve the accuracy of the classification model of random forest kidney disease,this paper proposes the use of multiple two-class random forests to solve the multi-classification problem of six kidney diseases.And through experiments,it is proved that the conjecture is feasible,and the optimal model of six kinds of nephropathy diagnosed by two-class random forest is designed.(4)The optimal model designed in(3)is used to analyze the needs of the nephropathy auxiliary diagnosis system,and the functional structure and role business of the system are designed.The system framework and the design and implementation of important modules in the system are introduced.Finally,the reliability and practicability of the system are verified by system performance verification.The renal disease auxiliary diagnosis system designed in this paper can be used to distinguish six kinds of kidney diseases,such as renal failure,pyelonephritis and nephrotic syndrome,and the accuracy rate is 87.7%.This system can help doctors to assist in the diagnosis of kidney disease,so that patients can be treated correctly as soon as possible to prevent the disease from worsening and leading to end-stage renal disease.