Prediction Of Acute Kidney Injury And Clinical Application Optimization Based On Machine Learning

Wisdom medical treatment uses computer technology,big data analysis technology,clinical diagnosis and treatment equipment,etc.,to make patients more scientific and intelligent diagnosis and treatment,thereby reducing the morbidity and mortality of patients.At present,big data,artificial intelligence,machine learning,etc.have gradually been applied in the field of medical servicess,and have produced enormous social and industrial space.Acute kidney injury is an emergency critical disease with extremely high morbidity and mortality.It explores the key risk factors affecting acute kidney injury,predicts and predicts the incidence of patients,optimizes the diagnosis and treatment process according to the patient’s disease chain and applies it to the clinic.Practice,it is of great significance for doctors to medically intervene in patients with acute kidney injury to reduce morbidity and mortality.This thesis based on the MIMIC-Ⅲ database,based on the basis of the 2012 Guidelines for Improving the Prognosis of Global Nephrology,the basis for early diagnosis of patients with acute kidney injury,1698 patients who met the conditions and corresponding basic personal information were extracted from the database.The indicator data is preprocessed in R language,mainly for pre-processing such as filling and discretization of missing values.The rough set,the grey correlation degree,the cellular automata and the genetic algorithm were organically combined,and the cell genetic algorithm of the subject intelligence screening was used to perform 20 independent repeated experiments,and finally 14 key risk factors affecting acute kidney injury were obtained.Then,using logistic regression,decisiosn tree-based adaptive reinforcement(Adaboost)algorithm,artificial neural network(MLP)to predict and predict acute kidney injury,the performance of the prediction model from the aspects of accuracy,accuracy and recall rate Conduct a comprehensive evaluation.The results show that the input of the key indicators based on the cellular genetic algorithm for prediction is better than the direct prediction by all the indicators not selected by the cellular genetic algorithm.The prediction effect of the integrated algorithm is better than that of the single algorithm.The performance of artificial neural network is better than decision tree based adaptive enhancement algorithm and logistic regression.Finally,by analyzing the variation of urine volume and creatinine in patients with acute kidney injury in emergency department,the value and clinical significance of urine volume and creatinine variability in patients with acute kidney injury are analyzed.From emergency data collection and information sharing,according to key indicators Early judgment of acute kidney injury and bedside monitoring system optimize the existing diagnosis and decision-making methods in several aspects of the clinician’s decision-making path to reduce the incidence and mortality of acute kidney injury.This paper has 37 pictures,18 tables,and 60 references.