基于机器学习的胆结石疾病预测效能研究OA

Objective To explore the performance differences of different machine learning models in the prediction of gallstone diseases,identify key predictive indicators,and provide data-driven intelligent solutions for early screening,early risk warning and clinical decision-making.Methods Based on 38 feature variables and one dichotomous target variable of 319 subjects in UCI gallstone database,the data were divided into training set and test set by stratified sampling method,and three prediction models of random forest,support vector machine(SVM)and extreme gradient boosting(XGBoost)were constructed.The performance of the models was comprehensively evaluated by accuracy,recall rate,precision rate and F1 value,and the key risk factors were identified by combining feature importance analysis.Results The study showed that all three models could effectively predict gallstones.Among them,the XGBoost model had the best performance,with an accuracy rate of 90.77％,a recall rate of 93.94％,a precision rate of 88.57％,and a F1 value of 91.18％,achieving the optimal balance between missed diagnosis and misdiagnosis risk.The random forest model demonstrated robust performance and could serve as an alternative,while the SVM model showed weaker adaptability and was not recommended for direct clinical screening.Feature analysis confirmed that C-reactive protein,vitamin D,and obesity degree were core indicators for gallstone prediction.Conclusion The XGBoost algorithm demonstrates significant advantages in processing high-dimensional clinical data and capturing complex feature interactions,making it suitable for gallstone disease prediction tasks.This study identifies the critical roles of inflammatory response and metabolic abnormalities in gallstone formation,providing a tool for early disease screening,facilitating the advancement of prevention strategies,and offering a methodological reference for the clinical translation of machine learning.