Abstract: Objective:To explore the clinical value of gradient boosting machine(GBM)model in predicting rebleeding in patients with non-variceal upper digestive bleeding (NVUDB).Methods:Clinical data of 258 patients with NVUDB admitted to our hospital from October 2020 to December 2021 were retrospectively analyzed, and the data set was randomly divided into training set and validation set according to the ratio of 7∶3, which were used to construct GBM model and verify the reliability of the model, respectively.Receiver operating characteristic(ROC)curve was used to analyze and evaluate the performance of the model, the calibration curve was used to evaluate the consistency between the model prediction probability and the sample probability, and the decision curve was used to evaluate the clinical practicability of the model.Results:The incidence of rebleeding in NVUDB patients was 20.9%.The top five important feature scores in the GBM algorithm model were Rockall score, shock on admission, D-dimer level, albumin level, and red blood cell distribution width.The area under the curve (AUC) of the training set was 0.985 (95% CI: 0.971-0.998), and the validation set was 0.873 (95% CI:0.785-0.960).The prediction accuracy of the training set was 92.2%, and the prediction accuracy of the validation set was 83.3%.The calibration curve showed that there was a good consistency between the predicted value of the GBM model and the actual observation value, and the model could predict the actual probability well.Decision curve analysis showed that the model had good clinical performance.Conclusion:The GBM algorithm model can better predict the risk factors of rebleeding in patients, and has high clinical effectiveness.