Prediction of vancomycin plasma concentration and adverse reactions based on ensemble learning
Objective To establish an ensemble learning model to predict the plasma concentration and adverse reactions of vancomycin,and to provide reference for its individualized medication.Methods The related data of patients from Changsha Hospital of Hunan Normal University from 2021 to 2023 were collected.Six machine learning methods,including Logistic Regression,Naive Bayes,Random Forest,Support Vector Machines,Gradient Boosting Decision Tree and Extreme Gradient Boosting,were used for the modeling.Meanwhile,the ensemble learning model was constructed to select the optimal subset to compare the prediction of models.Results Totally 205 cases were included,and the ensemble learning model based on the optimal subset predicted best.The root mean square error of this model in plasma concentration prediction was 7.703 and the mean absolute error was 6.492.The prediction accuracy of adverse reactions was 0.951,F1 score was 0.750,AUC was 0.959,and AUPR was 0.850.Conclusion The ensemble learning model based on the optimal subset can accurately predict the plasma concentration and adverse reactions of vancomycin,which provides a basis for precise individualized vancomycin anti-infection treatment effectively and safely.
NETL
NSTL
万方数据
