目的 探讨多种机器学习模型预测机器人辅助肾部分切除术（robot-assisted partial nephrectomy，RAPN）后肾功能减退的效能，为临床风险分层提供依据。 方法 回顾性纳入2019年1月至2023年12月重庆医科大学附属第一医院泌尿外科733例肾细胞癌（renal cell carcinoma，RCC）行RAPN患者的临床数据，整合人口学特征、实验室指标及围手术期参数，构建7种机器学习模型，采用Shapley加性解释（Shapley additive explanations，SHAP）方法解析关键预测因子，并通过受试者工作特征曲线下面积（receiver operating characteristic curve area under the curve，ROC-AUC）评估模型性能。 结果 随机森林模型预测效能最优（AUC=0.84）。SHAP分析显示，中性粒细胞/淋巴细胞比值、肿瘤直径、凝血酶原时间国际标准化比值、白细胞计数及术中出血量等因素对术后肾功能减退有明显影响。 结论 本研究为临床提供了一种潜在的预测工具，可帮助识别高风险患者并优化术后管理策略。

Objective To compare the efficacy of various machine learning models in predicting renal function decline after robot-assisted partial nephrectomy（RAPN），and to provide evidence for clinical risk stratification. Methods This study retrospectively included the clinical data of 733 patients with renal cell carcinoma undergoing RAPN at the Urology Department of The First Affiliated Hospital of Chongqing Medical University from January 2019 to December 2023. Demographic characteristics，laboratory indicators，and perioperative parameters were integrated to construct seven machine learning models. Key predictors were interpreted using Shapley additive explanations（SHAP）. Model performance was evaluated using the area under the receiver operating characteristic curve （AUC）. Results The random forest model demonstrated the best predictive performance（AUC=0.84）. SHAP analysis identified neutrophil-to-lymphocyte ratio，tumor diameter，the international normalized ratio of prothrombin time，white blood cell count，and intraoperative blood loss as significant factors influencing postoperative renal function decline. Conclusion This study provides a potential predictive tool for clinical practice，aiding in identifying high-risk patients and optimizing postoperative management strategies.