急性胰腺炎(acute pancreatitis, AP)是临床医生面临的常见疾病，其中重症急性胰腺炎（severe acute pancreatitis，SAP）病死率较高，因此早期识别可能发展为SAP的患者对于指导治疗意义重大。机器学习是一种多层描述的表征学习，从已有数据中分析挖掘获得规律，并利用这些规律对未知数据做出预测的算法。本研究基于机器学习建立了一个SAP预测评分系统，其可在入院24 h对患者SAP风险进行预测，预测准确率高达87.36%，AUC 94.11%。该模型可以更好的辅助临床决策和治疗，指导医生更早的进行相关干预。“,”Acute pancreatitis (AP) is a common disease faced by clinicians. Severe acute pancreatitis (SAP) has a high mortality rate, so early identification of patients who may develop into SAP is of great significance for guiding treatment. Machine learning is a multi-layer representational learning algorithm that analyzes and obtains laws from existing data and uses these laws to make predictions on unknown data. This study established an SAP prediction scoring system based on machine learning, which can predict the SAP risk of patients within 24 hours. The prediction accuracy rate is as high as 87.36% and AUC 94.11%. The model can better assist clinical decision-making and treatment, and guide doctors to make relevant interventions earlier.