目的: 本研究在医保基金监管面临新挑战的形势下,探讨如何利用机器学习技术提升对医保违规行为的识别精度与监管效率,以促进基金的稳定与可持续发展。方法: 本研究为有效识别复杂隐蔽的医保违规行为模式,采用机器学习分层分析方法,对四川省C市某三甲医院2023—2024年三个高就诊量病种（乳腺恶性肿瘤、冠心病、急性胰腺炎）患者的结构化与非结构化医保数据进行层级化处理与特征整合,分析潜在的医保违规风险情况。结果: 风险异常值高度集中于特定年龄与住院时长的交叉分组中,精准定位了需重点监控的高风险人群。异常费用结构呈现高占比（提示过度医疗）与低占比（提示服务不足）两种差异化模式,揭示了医保违规行为的隐蔽性与动态适应性。该方法为构建智能监测系统、实施精准风险预警及推动医保治理模式向事前预警、事中干预转型提供了有效的技术路径。结论: 本研究验证了分层分析方法结合机器学习技术在精准识别潜在医保违规风险中的有效性与应用价值。未来,应着力推进跨部门数据治理与深度整合,持续优化算法模型的解释性与适应性,加强复合型人才队伍建设,并紧密结合DRG/DIP等支付方式改革,迭代构建更加智能、高效、稳健的医保基金综合治理体系,从而系统性筑牢基金安全防线,推动医疗保障事业高质量可持续发展。

Objective: This study explores how to utilize machine learning technology to enhance the identification accuracy and regulatory efficiency of medical insurance violations, given the new challenges faced by medical insurance fund supervision, in order to promote the stability and sustainable development of the fund. Methods: To effectively identify complex and concealed patterns of medical insurance violations, this study adopts a hierarchical analysis method based on machine learning, processes and integrates structured and unstructured medical insurance data from patients with three high-volume disease entities (breast malignancy, coronary heart disease, and acute pancreatitis) in a tertiary hospital in City C, Sichuan Province from 2023 to 2024, and analyzes potential risks of medical insurance violations. Results: Risk outliers are highly concentrated in cross-groupings of specific ages and lengths of hospital stays, accurately identifying high-risk populations that require key monitoring. The abnormal cost structure exhibits two differentiated patterns: high proportion (indicating overtreatment) and low proportion (indicating insufficient service), revealing the concealment and dynamic adaptability of medical insurance violations. This method provides an effective technical path for constructing an intelligent monitoring system, implementing precise risk warnings, and promoting the transformation of medical insurance governance models toward prior warning and in-process intervention. Conclusion: This study verifies the effectiveness and application value of the hierarchical analysis method combined with machine learning technology in accurately identifying potential risks of medical insurance violations. In the future, efforts should be made to promote cross-departmental data governance and deep integration, continuously optimize the interpretability and adaptability of algorithm models, strengthen the construction of composite talent teams, and closely integrate with payment method reforms such as DRG/DIP, iteratively build a more intelligent, efficient, and robust comprehensive governance system for medical insurance funds. This will systematically strengthen the safety defense line of the fund and promote the high-quality and sustainable development of the medical security cause.