《保险研究》20190705-《大数据背景下的农险反欺诈检测：国际经验与技术选择》（贺娟、肖小勇、谭偲凤、陶建平）

［摘   要］中国农业保险市场中欺诈骗保等违法行为不容忽视,亟需运用数据挖掘技术提高农业保险发展质量。首先,本文对反欺诈检测的常用方法进行了梳理,包括异常值检测、聚类法、线性回归法、社会关系网络分析法等方法。其次,本文总结美国运用数据挖掘技术开展农业保险反欺诈检测的基本经验。美国利用以政府主导、研究机构参与的模式,开发出多种欺诈检测项目,为美国农业保险节约巨额资金。再者,基于国际经验,本文提出适用于中国农业保险反欺诈检测的相关性异常值检测法、合谋关系检测法和机器学习法。最后,为进一步推动数据挖掘技术在中国农业保险反欺诈检测中的运用,本文提出建立农业保险大数据库、建立数据挖掘合作平台、建立常态化的数据利用机制、以及培育和激励数据挖掘人才等建议。

［关键词］农业保险；数据挖掘；欺诈检测

［基金项目］华中农业大学自主科技创新基金项目(2662017QD011)；国家自然科学基金青年项目(71703049)。

［作者简介］贺娟,华中农业大学经济管理学院副教授，E-mail:hejuan@mail.hzau.edu.cn；肖小勇,华中农业大学经济管理学院讲师；谭偲凤,华中农业大学经济管理学院博士生；陶建平,华中农业大学经济管理学院教授。

Anti-fraud Detection Crop Insurance in the Era of Big Data:International Experience and Techniques Selection

HE Juan,XIAO Xiao-yong,TAN Cai-feng,TAO Jian-pin

Abstract：In the crop insurance market of China,fraud and other illegal acts have been repeatedly banned but are still prevalent. It is urgent to use data mining technology to develop high-quality crop insurance programs in China. The techniques used for fraud detection mainly include methods such as outlier detection,clustering,linear regression,social network analytics and so on. In order to carry out data mining technology for fraud detection in crop insurance,the United States has established a cooperative mode of government and research institutions to develop a variety of data mining projects,saving large amount of agricultural insurance indemnity annually. By learning from the international experiences and combining with China's real situation,this paper proposed the correlation outlier detection method,collusion detection method and machine learning method for fraud detection in China's crop insurance market. In order to further promote the application of data mining technology in China's crop insurance fraud detection,this paper advised to establish a data warehouse of crop insurance and a data mining cooperation platform,set up a mechanism on regularized usage of the data,and cultivate data mining talents.

Key words：crop insurance; data mining; fraud detection