饮茶对胃肠道疾病风险的双重作用：基于可解释机器学习与大语言模型的联合预测辅助模型

打开文本图片集

Abstract：ObjectiveToexplorethecorrelationofteaconsumptionwithrisksof gastrointestinal diseasesusingarisk prediction model integrating interpretable machine learning and alarge language model. Methods A survey was conducted amongthe patients undergoing both gastroscopy and 13C-urea breath testing at Gastrointestinal Endoscopy Center of Anxi HospitalofTraditional ChineseMedicine.Univariateanalysis wasperformedtodeterminethesuitabilityoffeatureselection. Thecolected data wererandomly divided into training and testing sets ina 7：3ratio.Support Vector Machine （SVM）， KNearestNeighbors（K），ogisticegression （R），domFrest （RF），EtremeGdientoosting（GB），ndal Network （DNN）were aplied toidentifythe bestclasifier forpredicting high-risk gastrointestinalconditions.Bayesian optimizationalgorithm wasused toobtain theoptimalhyperparametercombinationsforthe6models.AfterModelfiting，the interpretabilityof the best models wasanalyzed using SHapley AdditiveexPlanations （SHAP）.The DeepSeek-R1 base languagemodel wasfine-tunedwith gastrointestinal diseasedatasetand Chinese medicalonlineconsultation datatoobtain thefinalmodel.Results Thestudyincluded503 participants.Alltheselectedfeaturesshowedassciationwithgastrointestinal diseases， but only age exhibited a significant linear correlation （ （β=0.023， SE=0.008， t=2.942 1 P=0.003 ）.DNNmodel performed the bestwithagocacy（68），ei（.68）ae（.85），e，U（.74）.t were age，DOBvalue，andsmoking history.Thelargelanguage modelconstructed providedrecommendationsconsistent with thoseof professional physicians based on gastroscopyresults. Conclusion DNN model is efectiveforpredicting gastrointestinal diseaseriskandoffers reliablesupportforclinicalriskassessmentanddecision-makingregarding endoscopy Smoking cessation，moderate alcohol consumption，andreasonable teaintakemay helpprevent gastrointestinaldiseases. Keywords： gastrointestinal diseases; helicobacter pylori; machine learning; risk prediction; large language model

幽门螺杆菌 （H.p） 是一种定植于胃黏膜的微需氧螺旋杆菌，被世界卫生组织列为I类致癌原。（剩余13714字）