基于识别和多重分类的反洗钱系统  

Anti-money Laundering System Based on Identification and Multi-classification

作　　者：肖琨 王云[2] 张桂刚[2] XIAO Kun;WANG Yun;ZHANG Gui-gang(School of Information and Communication Engineering,Hubei University of Economics,Wuhan 430205,China;Institute of Automation,Chinese Academy of Sciences,Beijing 100190,China)

机构地区：[1]湖北经济学院信息与通信工程学院,武汉430205 [2]中国科学院自动化研究所,北京100190

出　　处：《小型微型计算机系统》2019年第10期2046-2051,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61872443)资助

摘　　要：反洗钱(AML)对于现代社会金融体系的健全具有重要意义,因洗钱与其他类型的犯罪活动密切相关,且涉及的资金数额巨大.本文旨在开发一种货币交易的可疑行为检测和分类系统.首先,根据货币交易过程中所表现出的不同特点,将与洗钱相关犯罪活动分为五类.然后在交易数据的基础上建立了用户档案,并从档案中提取出涉及个人和网络效应的特征.结合这两种特征,分别建立了两种基于监督学习方法的检测分类模型.结果表明,两种模型均具有较好的准确度和召回率以及良好的鲁棒性,可进一步调整,以供实际应用.最后,将两个模型串联起来,结果显示了相对较好的整体性能,以及验证了系统的可行性.Anti-money laundering( AML) is an important part of the integrity of financial system in modern society because of the huge amounts of money involved in and close relationships with other types of crimes. This paper aims to develop a suspicious behavior detection and categorization system based on money transaction data. For preparation work,the associated criminal activities were divided into five categories according to their different characteristics show n in the money transaction process. Then on the basis of the transaction data,a user profile was created and new features concerning both individual parties and network effect were extracted from the profile. With combined features,two models were developed for detection and classification using supervised learning methods separately. The results show good accuracy and recall rate. Meanwhile,the models display good robustness to further adjustment for practical use. Finally,two models were connected in series,and the result show s a relatively good overall performance and verifies the feasibility of the system.

关 键 词：反洗钱 特征提取 分类 监督学习 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]