Screening Factors Associated with Voriconazole Clearance in Elderly Patients Based on Machine Learning Feature Selection Algorithms
10.3870/j.issn.1004-0781.2025.06.027
- VernacularTitle:基于机器学习特征选择技术筛选老年患者伏立康唑清除影响因素
- Author:
Ke ZHAO
1
;
Hongxin YANG
1
;
Hao GUO
1
Author Information
1. 内蒙古自治区人民医院药学处,呼和浩特 010017
- Publication Type:Journal Article
- Keywords:
Voriconazole;
Machine learning;
Featureselection;
Clearance
- From:
Herald of Medicine
2025;44(6):998-1003
- CountryChina
- Language:Chinese
-
Abstract:
Objective To screen and identify significant influencing factors associated with voriconazole clearance using machine learning algorithms.Methods This study utilized a clinical dataset of elderly patients undergoing voriconazole treatment at our institution.We employed seven feature selection techniques from three categories of methods(filter,wrapper methodsand Embedded methods)to identify crucial influencing factors related to voriconazole clearance.Then six algorithms including Stochastic Gradient Descent Regression(SGDR),Lasso Regression(LassoR),Ridge Regression(RidgeR),Random Forest Regression(RFR),XGBoost Regression(XGBR),and Support Vector Regression(SVR),were applied to identify the best models by 5 fold cross-validation and the mean absolute error.SHAP values were employed to analyze the importance of influencing factors.Results The SVR model outperformed other regressors and were considered.The model was developed for seven selected features,which demonstrated good precision with an average relative error was 32.57%,and with approximately 57.89%of predicted values falling within a relative error range of±20%.The SHAP results revealed the importance ranking of influencing factors as ALT,ALP,PLT,administration method,CREA,TBIL,and PCT.Conclusion Feature selection techniques applied in machine learning can be utilized for the screening of influencing factors linked to voriconazole clearance.
