Prediction Of Heparin Dose During CRRT Surgery Based On Regression Model

With the improvement of human living standards,people pay more and more attention to the physical health.In recent years,the incidence of chronic kidney disease has become higher and higher,and Continuous Renal Replacement Therapy(CRRT)is the preferred method for treating this disease.CRRT surgical treatment replaces impaired renal function by continuous blood purification.In the continuous blood purification process,the main injection is heparin dose.At this stage,the dose of heparin is mainly determined according to the doctor's experience.This requires high doctor's experience and it is inevitable that misjudgment will occur,which will not only waste precious medical resources,but also it also exposes patients to unpredictable risks.Therefore,in order to reduce the doctor's subjective misjudgment,this paper uses a machine learning related algorithm to explore a regression model to predict the heparin dose in CRRT surgery,in order to help doctors make the right decision.The specific research contents are as follows:Firstly,perform data preprocessing for a given sample set.The exact same row data set is deduplicated,in order to reduce the time for learning the data set in the future;For missing value processing,the mean value filling method is used to interpolate the missing values;The data is standardized by using the z-score algorithm and range scaling algorithm.Secondly,for the data set after data preprocessing,the combination of genetic algorithm and random forest algorithm is used to extract features.For data feature extraction,the data are normalized by z-score algorithm and range scaling algorithm,and 11 experiments are performed respectively.Finally,18 important attributes are selected from 22 attributes for subsequent model prediction.For unbalanced data processing,a combined sampled method that combines EasyEnsemble algorithm and SMOTE algorithm is adopted.Finally,for the choice of the model,the regression model is adopted.Before the model training,the targeted variable,namely heparin dose,was optimized by taking ln transform,and the experimental comparison verifies that the heparin dose optimization could make the model fitting effect better.When training the model,this paper uses the Decision Tree Regressor model,the Support Vector Regressor model and the Gradient Boosting Regressor model to respectively experiment with the original dataset and the dataset processed by the EasyEnsemble algorithm and the SMOTE algorithm.By comparing the Mean Absolute Error,Mean Square Error and R squared of the evaluation indexes of the model,finally,this paperconcludes that after the unbalanced data set processing,the Gradient Boosting Regressor model is selected to predict the heparin dose in CRRT surgery.