Purpose/Significance To construct a risk prediction model for postoperative massive bleeding in nasopharyngeal carcinoma after radiotherapy, and to evaluate its predictive performance. Method/Process Inpatients with major bleeding after radiotherapy for nasopharyngeal cancer in the First Affiliated Hospital of Zhengzhou University from 2016 to 2019 are selected as the study objects, and the same number of patients without major bleeding are randomly selected as the control group. The medical record index data of the two groups of patients are collected, and various machine learning algorithms are applied respectively and the optimal algorithm is selected to build the model. Result/Conclusion The model based on support vector machine (SVM) algorithm has a recall rate of 0.94, an F1 value of 0.93, and a precision of 0.93, showing the best performance. It can be used to construct a prediction model for postoperative massive bleeding in nasopharyngeal carcinoma, and provide more accurate personalized prediction for patients, which has good clinical application prospects.