基于LightGBM的糖尿病风险预测方法研究

打开文本图片集

中图分类号：TP399 文献标识码：A文章编号：1006-8228（2026）01-23-04

Abstract：DiabetesisconsideredtobeoneofthemajorhealthcareproblemsaffectingmillonsofpeoplearoundtheworldAsthe prevalenceofthediseasecontinuestorise，researchershavebeenworkinghardtodevelopanacuratepredictionmodelfor diabetes.Inrecentyears，theresearchteamhasexploredtheaplicationofmachinelearinginthehealthcarefield.Inorderto predictdiabetesaccurately，thisstudyconductedexperimentsonsevenmachinelearmingalgorthmsonthePimadiabetesdataset， andusedexploratorydataanalysismethodstoidentifypaternsofthedataset'sfeatures.Inaddition，upsampling，normalization， featureselection，andhyperparametertuningareusedforpredictiveanalysis.Afteranalyzing theresultsusingvariousmachine learning metrics and k -fold cross-validation techniques，LightGBM achieved the highest classification accuracy of 98.2% among all clasifiers，demonstrating itsgeneral applicabilityto other diseases with similar pathological characteristics.

Keywords：LightGBM;Diabetes;Machine Learning;Risk Prediction

0引言

糖尿病是一个主要的全球健康问题，预计到2045年将影响全球 48% 的人口。（剩余5462字）