摘要Objective:To overcome the limitations of traditional single-indicator diagnosis for liver function abnormalities during pregnancy,this study aims to con-struct a machine learning prediction model based on bile acid metabolic profile data to enhance the accuracy of clinical diagnosis by analyzing bile acid metabolism.Methods:A total of 200 pregnant women who underwent prenatal examinations from January to December 2024 were selected and divided into a liver dysfunction group and a control group,with 100 cases in each group.Six indicators related to bile acid metabolism and liver function were extracted,including total bile acid(TBA)and gly-cocholic acid(GCA).The model was constructed using the randomForest package in R language,with parameters optimized via 10-fold cross-validation,and the dataset was split into a training set and a test set at a ratio of 7∶3.The model was evaluated using accuracy,sensitivity,and other metrics;the contribution of each indicator was ana-lyzed via variable importance,and inter-group differences in indicators were com-pared.Results:The random forest model achieved an accuracy of 89%,sensitivity of 87%,specificity of 91%,and AUC of 0.93 on the test set,which was significantly su-perior to traditional TBA-based diagnosis(accuracy:75%,sensitivity:72%,specifici-ty:78%,AUC:0.81,P<0.001).Feature importance analysis showed that TBA(40%),GCA(25%),and ALT(18%)were core factors,cumulatively contributing 83%.Inter-group comparisons showed that the six indicators,including TBA and GCA,were sig-nificantly higher in the abnormal group than in the control group(P<0.001),with TBA showing the most prominent difference.Conclusion:The model based on the random forest algorithm can effectively integrate multi-indicator correlations,accurately iden-tify abnormal liver function during pregnancy,provide a reliable tool for early clinical diagnosis,and help to improve maternal and infant health outcomes.