摘要
在基于机器学习的诸多应用中,采用的多是强监督学习方法,但这类方法对标签的依赖性极强。在一些应用场景中,标签的获取往往非常困难,如互联网金融中的网络借贷场景。为了解决这一问题,研究弱监督学习领域的相关算法,探索相关算法在网络借贷反欺诈方面的应用,并在一个银行的网络借贷数据集上实现了一个基于弱监督相关算法的反欺诈模型。结果显示,所采取的基于图的弱监督算法能够有效识别借贷欺诈。
In many applications based on machine learning,strongly supervised learning method is adopted,but this kind of method is highly dependent on tags.In some application scenarios,it is often very difficult to obtain tag,such as the online lending scenario in Internet finance.In response to this problem,the relevant algorithms in the field of weakly supervised learning are explored,the application of related algorithm in anti-fraud of online lending services discussed,and an anti-fraud model based on weak supervision related algorithms in a bank’s online lending data set implemented.The experiment results indicate that the graphbased weak supervision algorithm can effectively identify loan fraud.
作者
郑忠斌
胡瑞鑫
周宁静
王朝栋
ZHENG Zhong-bin;HU Rui-xin;ZHOU Ning-jing;WANG Chao-dong(Industrial Internet Innovation Center(Shanghai)Co.Ltd.,Shanghai 201303,China;Tongji University,Shanghai 200092,China)
出处
《通信技术》
2020年第10期2562-2566,共5页
Communications Technology
关键词
反欺诈
数据标签
弱监督学习
网络借贷
anti-fraud
data tag
weakly supervised learning
online lending service
