气象因素驱动的布地奈德高用药日预测集成学习模型的构建与比较

打开文本图片集

中图分类号R952 文献标志码A 文章编号1001-0408（2025）21-2723-04

DOI 10.6039/j.issn.1001-0408.2025.21.18

ABSTRACTOBJECTIVETo construct ensemblelearning modelsforpredicting high-usedaysof budesonide based on meteorologicalfactors，therebyprovidingreferenceforhospitalpharmacymanagement.METHODSMeteorologicaldatafor2024 andoutpatientbudesonideusagedatafromthejurisdictionofSanming HospitalofIntegratedTraditional ChineseandWestern Medicinewerecollcted.High-usedaysweredefinedasthe75thpercentileofoutpatientbudesonideusage，andacoresponding datasetwas established.Theprediction task was formulatedasaclasification problem，and threeensemblelearingmodels were developed：RandomForest，ExtremeGradient Bosting（XGBoost），andHistogram-based GradientBoostingClasifier.Model performancewasevaluatedusingaccuracy，precision，recall，F-score，andlog-lossModel interpretabilitywasanalyzedusing ShapleyAdditiveExplanations（SHAP）.RESULTSTheHistogram-basedGradientBoosting Classifierachieved thebest performance （accuracy ，Fl-score =0.48 ），followed by XGBoost （accuracy =0.74 ，Fl-score =0.43 ）and Random Forest （accuracy ⋅=0.72 ，F1-score ⋅=0.22 ）.SHAP results suggested that the prediction results of the last two models have the highest correction.CONCLUsIONSEnsemble learningmodelscanefectivelypredicthigh-usedaysof budesonide，withtheHistogrambasedGradient Bosting ClasifierdemonstratingthebestpredictiveperformanceLowtemperature，high humidity，andlow atmospheric pressure show significant positive impacts on the prediction of daily budesonide usage.

KEYWORDS budesonide；meteorological factors；ensemble learning；explainable artificial intelligence

呼吸系统疾病的发生与发展受多种因素影响，气象因素作为重要的环境因素之一，对患者的疾病进展及药物使用需求具有潜在影响[-2]。（剩余4415字）