目的 收集患者临床信息，采用机器学习算法构建患者静脉用药风险预测模型。 方法 回顾性纳入静脉用药患者（建模组1 302例和验证组281例），采用药学监护联盟协会提出的药物相关问题V 9.09分类标准分析患者存在的药物相关问题，采用logistics回归、神经网络、CHAID决策树、贝叶斯网络、支持向量机等机器学习算法构建静脉用药风险预测模型，并采用混淆矩阵格式对各预测模型进行评价。通过准确率、召回率、精确率、F1值以及生成验证受试者工作特征曲线下面积（area under curve，AUC）评价模型的预测性能。 结果 患者药物相关问题发生率为26.9%。患者药物相关问题主要集中在治疗安全性方面（n=556，94.9%），其次是治疗有效性方面（n=30，5.1%）。构建的模型中支持向量机的预测效能最好，AUC为0.826。 结论 机器学习算法构建的静脉用药风险预测模型预测效能良好，可为静脉用药安全管理提供新思路和新方法。

Objective To construct a predictive model for the risk of intravenous medication using machine learning algorithms based on the clinical information of patients. Methods A retrospective analysis was performed for the patients receiving intravenous medication，with 1302 patients in the modeling group and 281 in the validation group. The drug-related problem classification system V9.09 proposed by the European Society of Clinical Pharmacy was used to analyze the drug-related problems in patients. Machine learning algorithms，including logistic regression，neural network，CHAID decision tree，Bayesian network，and support vector machine，were used to construct risk predictive models for intravenous medication，and confusion matrices were used to evaluate the performance of each predictive model. Accuracy，recall rate，precision，and the area under the ROC curve（AUC） for the subjects in the validation group were used to evaluate the predictive performance of the model. Results The incidence rate of drug-related problems was 26.9% among these patients. These drug-related problems mainly involved treatment safety（n=556，94.9%），followed by treatment effectiveness（n=30，5.1%）. Among the models constructed，support vector machine algorithm showed the best predictive performance，with an AUC of 0.826. Conclusion The predictive model for the risk of intravenous medication constructed using machine learning algorithms has good predictive performance，which can provide new insights and methods for the management of intravenous medication safety.