基于“多层次分类”方法的异常P2P网贷借款识别

发布时间：2018-03-04 04:11

本文选题：PP网络借贷　切入点：异常检测　出处：《管理工程学报》2017年03期 　论文类型：期刊论文

【摘要】：随着互联网技术的发展,P2P网络借贷的用户与数据量与日俱增。识别出异常的借款标的,促进平台的健康发展一直是社会关注的热点与焦点。针对这一问题,本文提出了"多层次分类"方法,以lending club发布的交易数据为研究对象,分层次进行数据分析。在第一层次,首先采用基于密度的DBSCAN聚类算法,排除大量正常用户,减弱数据中正负两类分布不均衡的缺陷;在第二层次,采用一般分类算法进行分类,最终识别出平台的异常借款标的。数值实验发现,将"多层次分类"方法应用在P2P网络借贷中,相比于其他方法,能在保证分类器整体性能的情况下,更有效地识别出异常还款的借款标的。
[Abstract]:With the development of Internet technology, P2P network loan users and data are increasing. Identifying abnormal loan targets and promoting the healthy development of the platform has always been a hot spot and focus of social attention. In this paper, a "multi-level classification" method is proposed to analyze the transaction data published by lending club. In the first level, a density-based DBSCAN clustering algorithm is used to eliminate a large number of normal users. At the second level, the general classification algorithm is used to classify the abnormal loan target of the platform. Compared with other methods, the "multi-level classification" method is applied to P2P network lending, which can effectively identify the loan target of abnormal repayment under the condition of ensuring the overall performance of the classifier.
【作者单位】：厦门大学管理学院;浙江大学管理学院;
【基金】：国家自然科学基金资助项目(71572166);国家自然科学基金资助项目(71372057);国家自然科学基金资助项目(71301133) 厦门大学人文社科“校长基金-创新团队”基金资助项目(20720161044) 教育部人文社会科学基金资助项目(13YJC630033)
【分类号】：F713.36;F831.2

【相似文献】